torque sensor free power assisted wheelchair …237836/fulltext01.pdfthe project resulted in a...

Technical report, IDE0703, January 2007

Torque Sensor Free

Power Assisted Wheelchair

Master’s Thesis in Electrical Engineering

Jonas Johansson, Daniel Petersson

School of Information Science, Computer and Electrical Engineering

Halmstad University

Torque Sensor Free Power Assisted Wheelchair

Master’s Thesis in Electrical Engineering

School of Information Science, Computer and Electrical Engineering Halmstad University

Box 823, S-301 18 Halmstad, Sweden

January 2007

Description of cover page picture: The schematic principle of the torque sensor free power assisted wheelchair’s control system

i

Preface This master thesis was carried out at the electronic workshop at Halmstad University, Sweden, during the autumn and winter 2006. During the project, we have got great help from our main supervisor Ulf Holmberg, with the digital control part and from our second supervisor Björn Åstrand, who has supplied us with components and equipment. We are also grateful for the joystick controlled wheelchair that the company Decon Wheel, located in Hyltebruk, provided us with. We would also like to thank:

• Ruben Rydberg, at the electronic workshop, who gave us total access to the workshop, 24/7, and helped us on some occasions when hardware problem aroused during the project.

• Tommy Salomonsson, who helped us during the development phase of the development board “controller card 3.0”.

• Christer Gullbrand, who refreshed our knowledge about transistor theory.

• Håkan Pettersson, at the steel workshop, who constructed the mounting steel flange for the rotational encoders.

• Christopher Allen, who have read the thesis and corrected the English grammar. And finally, we would also like to thank our families and friends, who gave us support and motivation during the master thesis. Jonas Johansson & Daniel Petersson Halmstad University, January 2007

iii

Abstract A power assisted wheelchair combines human power, which is delivered by the arms through the pushrims, with electrical motors, which are powered by a battery. Today’s electric power assisted wheelchairs use force sensors to measure the torque exerted on the pushrims by the user. The force sensors in the pushrims are rather expensive and this approach also makes the wheels a little bit clumsy. The objective with this project is to find a new, better and cheaper solution that does not use expensive force sensors in the pushrims. The new power assisted wheelchair will instead only rely on its velocity, which is measured with rotational encoders, as feedback signal and thereby the project name “Torque Sensor Free Power Assisted Wheelchair”. The project consisted of two main parts; an extensive construction part, where an ordinary joystick controlled motorized wheelchair has been rebuild to the new power assisted wheelchair without torque sensors and a development part, where different torque sensor free controllers has been designed, simulated, programmed and tested. The project resulted in a torque sensor free power assisted wheelchair, where the final implemented design is a proportional derivative controller, which gives a very good assisting system that is robust and insensitive to measurement noise. The proportional derivative control design gives two adjustable parameters, which can be tuned to fit a certain user; one parameter is used to adjust the amplification of the user’s force and the other one is used to change the lasting time of the propulsion influence. Since the new assisting control system only relies on the velocity, the torque sensor free power assisted wheelchair will besides giving the user assisting power also give an assistant, which pushes the wheelchair, additional power. This is a big advantage compared to the pushrim activated one, where this benefit for the assistant is not possible.

v

Contents

Preface ..............................................................................................................................................i

Abstract ......................................................................................................................................... iii

1 Introduction ............................................................................................................................1 1.1 Power Assisted Wheelchair..............................................................................................1 1.2 Objective ..........................................................................................................................1 1.3 Strategy.............................................................................................................................2

2 Theoretical Background ........................................................................................................3 2.1 PAPAW............................................................................................................................3 2.2 Torque Sensor Free Solution............................................................................................4 2.3 Analysis ............................................................................................................................7 2.4 Design.............................................................................................................................12

3 Experimental Setup..............................................................................................................19 3.1 Original Joystick Controlled Wheelchair .......................................................................21 3.2 Modifications..................................................................................................................28

4 Experiments and Results .....................................................................................................33 4.1 Wheelchair Model ..........................................................................................................33 4.2 Controller Implementation .............................................................................................40

5 Conclusions ...........................................................................................................................51

References .....................................................................................................................................53

Appendix A - Controller Card 3.0 ..............................................................................................55

Appendix B - Block Scheme Simplification of Hori and Oh’s Controller ..............................63

Appendix C - Description of 22-pole Socket ..............................................................................65

Appendix D - Schematic Design of the External Middle Card ................................................67

Introduction

1

1 Introduction

For many years, there have only been three viable options for functionally impaired people, who are unable to propel an ordinary wheelchair. The three options are; using an electric powered wheelchair, driving a scooter or being pushed by an assistant. But recently, a new type of wheelchair has been developed that combines human power and electrical power. These are called power assisted wheelchairs.

1.1 Power Assisted Wheelchair

A power assisted wheelchair [1] combines human power, which is delivered by the arms through the pushrims, with electrical motors, which are powered by a battery. The power assisted wheelchair is aimed at customers, who have used a regular wheelchair for a long time, but who have become weaker or just need additional power when driving uphill. This kind of wheelchair will provide additional power for users, which will spare their wrists, elbows and shoulders.

1.2 Objective

Today’s electric power assisted wheelchairs use force sensors to measure the torque exerted on the pushrims by the user. This torque is then increased with the help of electrical motors. The wheelchair’s speed is also measured, with rotational encoders, to calculate how much power and for how long the power should be applied. The force sensors in the pushrims are rather expensive and this approach also makes the wheels a little bit clumsy. The objective with this project is to find a new, better and cheaper solution that does not use expensive force sensors in the pushrims. One idea proposed in a paper by Yoichi Hori and Schoon Oh [2] is the so called “sensor free” solution. In the paper the authors present a power assisting controller that does not use force sensors. The controller only relies on the rotational encoders that measure the speed. The name “sensor free” is misleading since speed sensors are used. In this case, “sensor free” only refers to the fact that no force sensors are used. Therefore the term “torque sensor free” will be used instead. The controller presented in the paper is made sensitive to disturbances, instead of rejecting disturbances as an ordinary controller. On the power assisted wheelchair, the force exerted by the human on the pushrims acts as a disturbance to the controller. The difficulty with this solution is to interpret the disturbance and to understand which disturbance movements that are caused by the human intervention (external disturbances). The main goals of this project are to construct a new kind of power assisted wheelchair that only relies on the speed sensors’ values and to implement, test and verify the “torque sensor free” solution.


2

1.3 Strategy

This project will be divided into two main parts; construction and controller development.

1.3.1 Construction

The construction part of the project will involve building an experimental setup with a motorized wheelchair. An ordinary motorized wheelchair, which is controlled by a joystick, will be used. This joystick controlled wheelchair will be rebuilt as a “torque sensor free” power assisted wheelchair. To be able to develop and program a new control system for the wheelchair, a development board will be constructed. This development board will be connected to the new wheelchair and control the wheelchair’s speed and direction. On the development board there will be a serial communication port, which makes it possible to monitor and change parameters, in the controller, in real time with the help of a computer. Since the new control system only will rely on the wheelchair’s velocity, speed sensors will be mounted on the wheelchair. The speed values from the sensors will be used as the only feedback signal to the controller system on the development board.

1.3.2 Controller Development

The controller development part of the project will consist of developing, programming and testing different torque sensor free control system on the new power assisted wheelchair. First, the control design presented by Hori and Oh will be simulated, tested and evaluated. Then, alternative control designs will be developed and implemented on the torque sensor free wheelchair. The stability criterion for all control designs will be investigated. The wheelchair, with the different control designs, will first be simulated on a computer, then tested in reality and finally a comparison between the simulated and real runs will be conducted.

Theoretical Background

3

2 Theoretical Background

This chapter is divided into four parts. In the first part, the theoretical background of the pushrim activated power assisted wheelchair’s different components and functions will be presented. In the next part, the main idea of the “torque sensor free” solution will be introduced. Then, the “torque sensor free” solution will be studied in the analysis part. Finally, alternative control structures for the power assisted wheelchair will be designed.

2.1 PAPAW

A Pushrim Activated Power Assisted Wheelchair (PAPAW) [1] is a combination of a manual wheelchair and a motorized wheelchair. The PAPAW is a new kind of wheelchair, where the force applied on each pushrim is measured with a torque sensor. A microcontroller is used to receive and interpret the torque signals from the two wheels. Then the microcontroller controls two DC-motors, one for each wheel, used to drive the wheelchair, depending on the applied force. The main idea of this kind of wheelchair is to amplify the human force by measuring the torque on the pushrims and use the electrical motors to assist the user with more power and move him further on each push. For example when the user travels uphill, the PAPAW will provide him with even more additional power, to help him up. Most PAPAWs have many intelligent inbuilt functions. One example of an inbuilt function [3] is, if the user has different levels of strength in his right and left arm, the PAPAW can adjust the support on the right and left wheel, to make the wheelchair go straight after all. Another example of an innovative function [4] is that the wheelchair keeps going straight despite disturbances in the lateral direction or when just driving with one hand. An improved control system for the PAPAW that sense if the user wants to turn or go straight ahead is presented in [5]. Furthermore, there exist functions that prevent the wheelchair from overturn [6], especially when driving uphill. A schematic example of the pushrim activated power assisted wheelchair’s control hardware is shown in Figure 2-1.

Microcontroller

Motor

Gearbox Gearbox

Wheel Wheel

Load Load

Motor

Speed Sensor

Torque Sensor Torque Sensor

Pushrim Pushrim

Speed Sensor

Figure 2-1. The schematic example of the PAPAW’s control hardware


4

A torque sensor is mounted on each pushrim to measure the torque applied by the user. The signals from the torque sensors are processed by the microcontroller that contains the embedded control system, which does all the real time calculations. The microcontroller controls the rotational speed and direction of the two motors. Each motor then drives a gearbox which in turn drives the wheel. The rotational speed on each wheel is measured with a speed sensor, where the signals from the sensors are used as feedback to the control system in the microcontroller. The influence of different loads on the wheelchair, like friction and weight, affect the motor’s torque and speed.

2.2 Torque Sensor Free Solution

Yoichi Hori and Schoon Oh [2] have presented a so-called “sensor free” solution based on velocity control. The authors have designed a controller that is sensitive to disturbances, which can be used in power assisted systems. Usually a power assisted system uses sensors, in most cases torque sensors, to measure the external forces caused by for example human influence. These systems that use force sensors are expensive and therefore a cheaper solution is desirable. In the paper, Hori and Oh have presented a power assisting controller that does not use these expensive force sensors. The new controller only relies on the external changes in the feedback signal. In other words, the controller interprets and amplifies the external disturbances, compared to an ordinary control system that suppresses disturbances. In the case with the pushrim activated power assisted wheelchair, the expensive torque sensors in the pushrims are used as a feedforward signal to the control system. With a velocity disturbance observed controller, these torque sensors can theoretically be removed. The wheelchair’s new control system will only rely on the wheel’s velocity values, when controlling the electrical motors. A schematic hardware example of this theoretically power assisted wheelchair with a velocity disturbance observed controller is shown in Figure 2-2.

Microcontroller

Motor

Gearbox Gearbox

Wheel Wheel

Load Load

Motor

Speed Sensor Speed Sensor

The difference between this new power assisted wheelchair and the pushrim activated one, in Figure 2-1, is that this does not use torque sensors. The embedded control system in the microcontroller has to be more advanced, since the control system only relies on the speed sensor values.

Figure 2-2. The schematic hardware example of the power assisted wheelchair with a velocity disturbance observed controller


5

2.2.1 Wheelchair Model

To be able to design a control system for the power assisted wheelchair, a model of the wheelchair has to be created. Hori and Oh present a wheelchair model with only one time-constant, where Jr is the inertia and Br is the damping factor. In Figure 2-3, the step response of the wheelchair model is shown. For example, a higher inertia Jr, results in a slower response and a higher damping factor Br, results in a lower speed of the wheelchair.

rr BsJ +1

y u

step

r

r

B

J

y

t

63 %

The wheelchair model can be obtained by exciting the motorized wheelchair system with different input signals. The response of the system can be seen, by measuring the output signal of the system, in this case the wheelchair’s velocity. If the motorized wheelchair is excited with a step signal to the motors, a step response of the wheelchair’s velocity will be shown in the measured data. For example, if the motorized wheelchair acts like a first order system, as shown in Figure 2-3, the parameters Jr and Br can be calculated with the help of the measured velocity.

2.2.2 The Control Design by Hori and Oh

Hori and Oh propose a controller design of the velocity disturbance controlled system. The proposed system, which is shown in Figure 2-4, consists of five blocks (transfer operators).

rr BsJ +1

Human Force (External Disturbance)

d

u

A

1++

sT

BsJ

a

rr 1

1

+sTa

MM BsJ +1

u

u + d

h ≈ d

Process (Real Wheelchair)

Desired Wheelchair

Desired Velocity

rv

rvMv

≈ u ≈ u + d

Real Velocity B1.

B2. B3.

B4.

B5.

Figure 2-4. Block diagram of the control design by Hori and Oh

Figure 2-3. The step response of the wheelchair model, where Jr is the inertia and Br is the damping factor.


6

The first block B1 is the process that is to be controlled, in this case the motorized wheelchair. The input signal to the process is the control signal u together with the human’s force (external disturbance) d and the output signal from the process is the wheelchair’s real velocity vr. The B2 block removes the wheelchair’s dynamics and outputs the process’ input signal that includes the external disturbance d. Since the degree of the nominator can not be higher than the degree of the denominator, Hori and Oh have added a polynomial Tas+1 in the denominator. As a result of that, they also added a third block B3 that contains the same polynomial in the denominator, which has the control signal u as input. The only disadvantage of the blocks containing the polynomial Tas+1 is that it will cause a small delay of its input signal. The point h in the figure is the difference between the B2 and B3’s output signals, which theoretically only contains the external disturbance d. The external force by the human at the point h is calculated below.

0,1

1

1

1)(

1

1

1

1)(

1

1

1

1

1)(

1

→→+

=

+−+

+=

+−+

+⋅

++=

+−

++=+

+=

aa

aa

arra

rr

ar

a

rr

rrr

TddsT

h

usT

dusT

h

usT

duBsJsT

BsJh

usT

vsT

BsJhdu

BsJv

Q

If the time-constant Ta is chosen small, the signal at point h will almost be equal to the disturbance d, since the delay will be neglectable. The disturbance signal at the point h then passes a fourth block B4 that contains the dynamics of a desired model of the wheelchair. B4 outputs a desired velocity vM. Then the difference between the desired velocity vM and the real velocity vr is multiplied by a constant A before the signal is fed back to the process as the control signal u. The effect of the control system can be altered by changing A. By setting A equal to zero, the controller will be turned off and the system will be in open-loop mode, which means that the system only will contain the process. The main idea of the Hori and Oh controller system, is that it will send a feedback signal u to the process (real wheelchair), which regulates the speed so the desired velocity vM is equal to the real velocity vr. In other words, the dynamics of the power assisted wheelchair can be modified, by just changing JM and BM in the desired wheelchair model block. However, the controller will only work correctly if the parameters Jr and Br of the real wheelchair are precisely calculated and remain the same during the drive.


7

2.3 Analysis

The “torque sensor free” controller by Hori and Oh, presented in Chapter 2.2, will be analysed in this section. First the controller will be discretized. After that, the disturbance to output sensitivity for the controller will be calculated. Finally, stability criteria that constrain the different parameters in the controller will be studied.

2.3.1 Discretization of the Control Design by Hori and Oh

Discretization is an important step towards making a continuous-time system implementable on a processor (computer). Therefore, the control design by Hori and Oh, mentioned in Chapter 2.2.2, will be discretized. The discrete-time form of the control design by Hori and Oh is shown in Figure 2-5.

)()(

1

1

−

−

qA

qB

Human Force (External Disturbance)

d

u

K

1

1)(

b

qA −

1−q

)( 1−qA

b

m

m

)( duA

Bvr +=

Real Velocity

B1.

B2. B3.

B4. B5.

dA

Bv

m

mm =

u

u + d

Process

uq 1− )(1 duq +−

dq 1−

rv

Desired Wheelchair

Desired Velocity

The discrete-time wheelchair process in block B1 is P(q

-1)=B(q -1)/A(q

-1), where A(q -1) and B(q

-1) are the process polynomials in backward-shift form. The backward-shift operator q

-1 is defined as q

-1y(k)=y(k - 1), where y(k) is the sample value at a given sample position k and y(k - 1) is the previous sample value. The discrete-time process P(q

-1) can be calculated by transforming the continuous-time wheelchair model (process) P(s) = 1/(Jrs + Br), presented in Chapter 2.2.1. The transformation can be done by using the formula for sampling of first order system.

( ) aB

bea

qa

qb

qA

qBqP

BsJsP

r

hJ

B

rr

r

r

111

11

11

1

11

11

,

1)(

)()(

1)(

+=−=

+==→

+=

−

−

−

−

−−

where the parameter a1 is the time constant, b1 is the gain factor and h is the sampling period.

Figure 2-5. Block diagram in discrete-time form of the proposed control design


8

The second block B2 contains an elimination function A(q -1)/b1 that removes the wheelchair’s

dynamics in the real velocity output signal vr. The resulting output signal from B2 will then be delayed one sample and only contain the control signal u with the external disturbance d. The block B2’s output signal is calculated below.

( ) ( )

( ) ( )duqduA

qb

b

Av

b

A

duA

qbdu

A

Bv

r

r

+=+⋅=⋅

+=+=

−−

−

11

1

11

11

The block B3 only contains a delay factor just like the continuous-time version. This factor in B3 delays the control signal u one sample, which results in the output signal q

-1u. The difference between the output signals from the blocks B2 and B3 will then only be the external disturbance, which is delayed one sample.

( ) dquqduq 111 −−− =−+ The external disturbance signal is fed to the fourth block B4, which is a model of the desired wheelchair. This model can be modified by changing the time constant am and the gain factor bm. B4 outputs the desired velocity vm.

dqa

qbdq

A

bv

m

m

m

mm 1

11

1 −

−−

+==

For example, by changing the desired gain factor bm to a greater value than the real gain factor b1, the desired velocity vm can be higher than the real velocity vr, when the system is exposed to an accelerating external disturbance d. By changing the desired time constant am to a greater value (closer to -1) than the real time constant a1, the time response of the desired velocity can be slowed down compared to the real velocity. This results in a longer lasting disturbance response. Then the difference between the desired and the real velocity passes an amplifier block B5. The constant, which was called A in the continuous-time version, is now called K since A is occupied by the polynomial A(z

-1) in the process. The constant K is used to regulate the influence of the velocity error between the real velocity vr and the desired velocity vm.

)( rm vvKu −=

For example, the controller in the wheelchair system can be turned off, by setting K to zero.


9

2.3.2 Block Scheme Simplification of the Control Design by Hori and Oh

The controller’s transfer function for the discrete-time version of the control design by Hori and Oh, needs to be calculated in order to simulate different parameter values, check stability for the process, etcetera. The controller’s transfer function can be obtained by using a block scheme simplification procedure. In this case, the discrete-time block scheme of the Hori and Oh control design will be reduced to only two blocks; the process block B(q

-1)/A(q -1) and the controller block

-S(q -1)/R(q

-1), which is shown in Figure 2-6. In Hori and Oh’s control system, the reference signal r is equal to zero, since the system does not use any feedforward signals. The output signal y is in this case the system’s velocity vr and d is the external disturbance caused by the human. The block scheme reduction of the discrete-time control design by Hori and Oh, seen in Figure 2-5, resulted in a new scheme, which is shown in Figure 2-7, where the controller transfer function is:

KBA

Kb

bAA

R

S

mm

mm

+

−

= 1

The R and S polynomial in the controller are then

KBAR mm += ( )KbbAAS mm )( 1−=

which results in the characteristic closed polynomial

( ) )()()( 1 KBAAKbbAABKBAABSARA mmmmmC +=−++=+= d

y

A

B 0=r u

( )( )KBA

KbbAA

mm

mm

+−− 1

All steps of the block scheme reduction procedure can be seen in Appendix B.

Figure 2-6. The closed loop block scheme with the process and the controller

Figure 2-7. The simplified block scheme of the control design by Hori and Oh

d

yr u

R

S−

A

B


10

2.3.3 Disturbance to Output Sensitivity Function

The output sensitivity function Sd can be used to study the controller’s sensitivity, when the system is exposed to disturbances. In the case with the velocity disturbance observed controller, the response of the control system can be examined, when different external disturbances are fed into the system. The output sensitivity function for the velocity disturbance observed control system can be calculated in the following way. The output sensitivity function affects the output signal y depending on the disturbance signal d.

dSy d=

In polynomial form the process and the controller are:

)( duA

By += y

R

Su −=

These polynomial equations can be added together to form a new function where the output signal y changes with respect to the disturbance d.

d

ABSAR

BRd

AR

BSA

B

ydA

B

AR

BSydy

R

S

A

By

C

43421 +=

+=⇒=

+⇒

+−=1

1

This new function is the sensitivity function Sd.

Cd A

BRS =

In the case with the control design by Hori and Oh, the sensitivity function will be:

( ))( KBAA

KBAB

A

BRS

m

mm

Cd +

+==

2.3.4 Stability

The stability of a system is a factor that is important to consider, when designing a controller. Stability means, if the system is not exposed to any input signals, the system’s output signal will gradually go to zero.


11

A system is stable if and only if the output signal y(k)→ 0 when k →∞ for all initial values in y. Consider the system below, with no input signals and where the initial condition differs from zero.

0,0)()( 1 >=− kkyqA For example, consider the first order system A(q -1) = 1 - λ q -1 where λ is the pole. This gives a general solution:

cky

ccy

ccy

cy

cy

kyky

kyky

kyqkyqA

kλλλλλλλ

λ

λλ

λ

===

==

==

−==−−

=−= −−

)(

)()3(

)()2(

)1(

)0(

)1()(

0)1()(

0)()1()()(

32

2

11

Q

To satisfy the stability criteria where y(k)→ 0 when k →∞ for all initial values c, the pole has to be inside the unit circle, in other words 1<λ .

The general solution can also be applied to higher order systems, which gives the criteria 1<iλ .

)1)...(1)(1)(1()1()( 113

12

11

1

11 −−−−

=

−− −−−−=−= ∏ qqqqqqA m

m

i

vi

i λλλλλ

In the case with the control design by Hori and Oh, the wheelchair process will be stable if the time constant a1 is inside the unit circle since a1 = - λ.

11)(

)()(

11

11

1

11 <−=

+== −

−

−

−− λλ a

qa

qb

qA

qBqP

The most important stability criteria to consider are the ones in the closed loop system. The closed loop system can be stable although the controller is unstable. It is therefore important to analyse the stability criteria for the output sensitivity function Sd, which is the closed loop system of the velocity disturbance observed control system. In the case with Hori and Oh’s control system, the stability criteria can be found in the characteristic closed polynomial Ac.

( ) ( )( ) ( ) ( )( )11

111

11

1 1111)( −−−−− +++=+++=+= qKaqaqKbqaqaKBAAA mmmc

The closed loop system has the two stability criteria 12,1 <λ where ma−=1λ and ( )112 Kba +−=λ .


12

2.4 Design

In this section, new velocity disturbance observed controllers for the power assisted wheelchair will be designed. First, a general structure of a velocity disturbance observed control system will be presented in Chapter 2.4.1. With this general structure, new kinds of velocity disturbance observed control systems can be designed by choosing four different parameters. After that, in Chapter 2.4.2, two new control designs will be proposed for the power assisted wheelchair. The first one is a regular proportional control design, which has a positive feedback instead of a negative feedback. The second one is a proportional derivative control design, which also has a positive feedback.

2.4.1 General Structures

The controller –S/R for a velocity disturbance observed control system can be written in a general form, where S = S0 + QA and R = R0 + QB. This general controller, see Figure 2-8, contains a process eliminating part A/B that removes the process’ influence of the control signal containing the external disturbance.

uyy

A

B du +

Process General Controller

QBR

QAS

−+−

0

0

The transfer operator Q, in the general controller, has the polynomial Qn in the nominator and the polynomial Qd in the denominator. These two polynomials can be chosen freely, as long as the transfer operator Q is stable. Therefore, the general controller polynomials will be:

AQSQS nd += 0

BQRQR nd −= 0

The general characteristic closed polynomial will then be:

)()()( 0000 BSARQAQSQBBQRQABSARA dndndC +=++−=+=

which will give the general output sensitivity function:

)( 00 BSARQ

BR

A

BRS

dCd +

==

Figure 2-8. The process block and the general controller block that has a process eliminating function of the y signal


13

By choosing the polynomials S0, R0, Qn and Qd, different structures of velocity disturbance observed controllers can be designed. For example, the control design by Hori and Oh can be obtained if the S0, R0, Qn and Qd, in the general disturbance observed controller are chosen to:

( ) ( )( )

+=−−⋅=−=−+=

⇒

=

−=

=

=

KBABKbbAR

KbbAAAKbbKAS

AbQ

Kb

bQ

R

KS

mMmM

mMmM

Md

mn )(1

)()(1

1

11

1

1

0

0

The output sensitivity function of the control design by Hori and Oh is calculated, with the general output sensitivity function, to:

( ))()( 00 KBAA

KBAB

BSARQ

BR

A

BRS

M

mM

dCd +

+=+

==

2.4.2 Proposed structures

Two new velocity disturbance observed controllers will be designed. The first one is the proportional control design and the second one is the proportional derivative control design.

Proportional Control Design

A proportional controller, which is shown in Figure 2-9, will give the simplest kind of velocity disturbance observed controller. In regular cases, the constant K in the proportional controller is chosen positive, which will give a negative feedback that will reject the external disturbance d. But, in the case with the velocity disturbance observed design, the constant K is chosen negative instead of positive, resulting in an amplification of the external disturbance instead of a rejection of the disturbance. This proportional controller will give the user a satisfying support (additional power). But, the proportional controller will have a limitation of how much the initial force applied on the pushrims can be amplified. Further details about the proportional controller can be found in Chapter 4.2.1.

d

y

A

B

K−

0=r u

Figure 2-9. The proportional control system


14

The proportional controller can also be obtained if the polynomials in the general disturbance observed controller, mentioned in Chapter 2.4.1, are chosen to be:

=−=<=+=

⇒

===

1

0

0

10

00

0

QBRR

KKQASS

Q

R

KS

The output sensitivity function for the proportional controller will then be

KBA

B

BSAR

BR

A

BRS

Cd +

=+

==

The stability criteria for the closed loop system can be found in the characteristic closed polynomial Ac.

111

11

11 )(1)1( −−− ++=++=+= qKbaqKbqaKBAAc

The closed loop system with the proportional control design has the stability criteria 11 <λ where

( )111 Kba +−=λ . With known values of the time constant a1 and the gain factor b1, the limits of the gain constant K can be calculated. The stability criterion together with K < 0 will give the limits

( ) ( )( )

01

0

11

1

1

1

1

1

11

1

1

1

1

1

1

1

1

11

1

1

11

11

1111

<<−−⇒<

−<<−−

−=

=+−−=

−=+

=+−−=+−

⇒<+−

Kb

aK

b

aK

b

a

b

aK

Kba

b

aK

Kba

Kba

KbaKba

Proportional Derivative Control Design

To be able to increase the initial force, a derivative part is added to the proportional controller. This results in a proportional derivate (PD) control design. The wheelchair’s velocity is the only information that is used in the controller, but additional information can be obtained from the velocity by differentiating it. The derivate of velocity is acceleration and by looking at the acceleration, information about the initial force caused by the user can be found. Thus, a PD-controller can be used to adjust and increase the initial force.


15

The PD-controller, shown in Figure 2-10, contains two gain constants; the proportional constant Kp and the derivative constant Kd.

u y

⋅+−dt

dKK dp

The derivative part in the PD-controller can not be implemented in this way:

)()( kydt

dKkD d ⋅=

Since the velocity signal y(k) can be discontinuous and noisy, the signal y(k) should not be differentiated. That would result in very high spikes in the differentiated velocity, in other words in the acceleration. The derivative part is instead implemented as a high pass filter with a limited noise feedback, which reduces the high-frequency gain. In continuous-time the filtered derivative part will be

)()( sYKsD dd=

where Yd is the filtered acceleration, which is:

)(

1

)( sY

sN

Ks

sYd

d

+=

The positive constant N in the filter is used to alter the high-frequency gain limit. A higher value of N will result in a less filtered (noisier) signal and a lower value will result in a lower limit in the frequency gain curve, shown in Figure 2-11.

ω

)( ωiY

Yd N

s

Figure 2-10. The proportional derivative controller

Figure 2-11. The frequency gain curve where N limits the high-frequency gain


16

This filter can be written in discrete-time form by using backward difference:

h

qs

11 −−=

where h is the sampling period. The filter then becomes:

( )( )

( ) )(1

)()(1

1

)(

)(

11

1

)()(

1

)(

1

1

1

1

1

1

kyqKNhK

qNkyky

NhqKh

qNh

ky

ky

h

q

N

Kh

q

kysY

sN

Ks

sY

ddd

dd

dd

dd

−

−

−

−

−

−

−+−=⇒

+−

−

=

+−

−

=⇒

+=

( ))(

1

1

)(1

1

ky

qNhK

K

qNhK

N

ky

d

d

dd

−

−

+−

−+

=Q

The PD-controller with the filtered derivative (acceleration) will then become:

)()()( kyKkyKku ddp −−= ( )

+=

+=

−−= −

−

NhK

Kc

NhK

Nc

kyqc

qcky

d

d

d

d

2

1

12

11 ),(1

1)(

( ) ( ) ( ))(

1

11)()(

1

1)(

12

11

12

12

11 ky

qc

qcKqcKkuky

qc

qcKKku dp

dp

−−+−

−=⇒

−−+−= −

−−

−

−

( ) ( ))(

1)(

12

1211 ky

qc

qcKcKcKK

R

Sku pddp

−

−

−−−++

−=−=Q


17

The control system with the proportional derivative controller including the limitation filter is shown in Figure 2-12.

d

y

A

B 0=r u

( ) ( )1

2

1211

1 −

−

−

−−++−

qc

qcKcKcKK pddp

The stability criteria for the closed loop system can be found in the characteristic closed polynomial Ac.

( )( ) ( ) ( )( )( ) ( ) ( )

( ) ( ) 2

2

2111211

1

11121

2211

111

221

121

1211

11

12

11

1

1

11

−−

−−−−

−−−−

−−−+++−+=

−−+++−−+=

−−+++−+=+=

q

g

cKbcKbcaq

g

cKbKbcaA

qcKcKbqcKKbqcaqcaA

qcKcKcKKqbqcqaBSARA

pddpc

pddpc

pddpc

4444 34444 214444 34444 21Q

Since Ac is a second order system, the stability limits of the parameters g1 and g2 can be described by a stability figure, shown in Figure 2-13, where the parameters should be inside the enclosed area.

g2

g1

g2 = 1

g2 = -g1 - 1 g2 = g1 - 1

The stability criterion for the proportional derivative controller is then:

( )( )

<−>

−−>

−−−=++−=

1

1

1

2

12

12

2111212

111211

g

gg

gg

cKbcKbcag

cKbKbcag

pd

dp

Figure 2-12. The proportional derivative control system

Figure 2-13. The stability area for the parameters g1 and g2


18

Experimental Setup

19

3 Experimental Setup

To be able to test and verify this torque sensor free control system, an experimental setup had to be constructed. The company Decon Wheel [7] supplied this project with an ordinary joystick controlled motorized wheelchair, which had to be rebuilt to a power assisted wheelchair with a control system that only relies on velocity disturbance observation. In Figure 3-1, a schematic example of the original joystick controlled wheelchair’s control hardware is shown.

Joystick Module

Battery Motor

Brake Brake

Gearbox Gearbox

Wheel Wheel

Load Load

Motor

External Processor Card

Serial Communication

Power Module

Main Power Card

The user, driving the motorized wheelchair, controls the wheelchair with a joystick module. The joystick module sends the user’s commands to a power module via a serial communications protocol. The power module consists of two cards; a main power card that holds all power electronic devices and an external processor card, which handles the communication between the joystick module and the main power card. The power module, which is supplied by a 24 volts battery, translates the received commands from the joystick module into control signals (power signals) that drive the two electrical motors on the wheelchair. Each motor is connected to a gearbox, with a resulting gear ratio of 81:1, which increases the motor’s torque and decreases the motor’s rotational speed proportionally. Each gearbox is in turn connected to a wheel with a pushrim, which also can be used to propel the wheelchair. Each wheel is exposed to different loads, e.g. weight, friction or up- and down-hill driving, that affects the motor’s torque. On each motor there is also a solenoid brake mounted, which is used to brake the motor axis when the power module gets a braking command from the joystick module.

Figure 3-1. Schematic example of the joystick controlled wheelchair’s control hardware


20

A schematic example of the new modified power assisted wheelchair’s control hardware is illustrated in Figure 3-2. All significant modifications of the control hardware in the figure are highlighted in grey.

Controller Card

Battery

Gearbox Gearbox

Wheel Wheel

Load Load Motor

External Middle Card

Modified Power Module

Main Power Card

Rotary Encoder Rotary Encoder Motor

“Direct Connected”

The joystick module has now been removed and replaced by a programmable controller card that will be used to develop the new control system. The controller card is “directly connected” to the power module’s hardware instead of using the serial communication. The external processor card, which was placed inside the power module and handled the serial communication protocol translation, has been replaced with an external middle card that translates control signals from the controller card before the signals enter the main power card. The external middle card also works as a safety card that protects the power module from being destroyed by, for example, careless coding of the control system. The solenoid brakes, placed on each electrical motor, were removed to be able to mount rotary encoders instead. The speed values from the rotary encoders will be used as feedback signals to the control system. This chapter has been divided into two parts. The first part, Chapter 3.1, explains the different components and functions of the original joystick controlled wheelchair. The second part, Chapter 3.2, describes all modifications done to rebuild the original wheelchair to the disturbance controlled power assisted wheelchair.

Figure 3-2. Schematic example of the velocity disturbance controlled power assisted wheelchair’s control hardware

Experimental Setup

21

3.1 Original Joystick Controlled Wheelchair

The joystick controlled wheelchair, shown in Figure 3-3, which has been used for the project experiments, has been provided by the company Decon Wheel [7]. The wheelchair is an electric powered wheelchair, with one electrical DC-motor on each wheel. The DC-motors are controlled and powered by a power module. The power module receives its supply from a 24 volts battery. The user, which drives the motorized wheelchair, controls the wheelchair with the help of a joystick module that communicates with the power module by a serial protocol. The controller in the power module is very simple, and only follows the commands from the joystick. The wheelchair itself is constructed by Be Rolka aktiv [8] and Decon Wheel has replaced the wheels and mounted their own so-called spiderwheels. The spiderwheels are made of magnesium and the rims are made of aluminium, which makes them lightweight. The wheelchair has also the ability to decouple the motors with a lever and the electrical wheelchair will then become a regular wheelchair, which can be propelled forward using the pushrims.

3.1.1 Joystick Module

The motorized wheelchair is controlled by a joystick module manufactured by the company PG Drives Technology [9], shown in Figure 3-4. This joystick model is a version without lightning control. The joystick module can display the wheelchair’s status on a colour bar graph display. The display can for example show the battery state and also if there are some technical errors on the wheelchair like disconnected motors, solenoid brake faults or power module faults. There are three buttons on the joystick module; one on/off switch, one horn switch and one switch that is used to change speed-modes. The joystick module is connected to the wheelchair’s power module; these components communicate with each other using a serial communication protocol. The speed and steering commands from the user are converted into data packets by the joystick module’s internal processor, which are then sent to the power module via the serial communication. See section External Processor Card in this chapter for further details about the communication protocol.

Figure 3-3. The joystick controlled wheelchair

Figure 3-4. The joystick module


22

3.1.2 Power Module

The power module manufactured by the company PG Drives Technology [9], shown in Figure 3-5, is the main system of the electrical wheelchair. The power module, which is supplied by a battery, receives the speed and steering commands from the joystick module and translates the commands into high current signals to the two motors. The power module has five outlets in the front, seen in the figure. From left to right there is a connection to the external joystick module, the left motor, the 24 volt battery, the right motor and finally a connection for optional devices like, for example, a lightning module. Each motor outlet has also a connection to a magnetic brake that is mounted on each motor axis, which makes it possible for the module to brake the wheelchair. The power module is able to deliver a very high current per electrical motor, despite its small size. The power module uses a passive cooling solution. The case, enclosing the hardware, is made of aluminium with fins, which spread the heat generated by the components inside the box. Inside the power module, seen in Figure 3-6, there are two PCB (Printed Circuit Board) cards with associated components. There is one main power card, which contains the power electronic devices, like motor bridges, high voltage transistors, large electrolytic capacitors and a mechanical relay. The other circuit board is an external processor card, which is connected to the main board through a pin socket. On the external card there is a microprocessor that controls the main power card and handles the communication with the external devices.

Figure 3-6. The power module in pieces. The two PCB cards inside the power module are the main power card and the external processor card.

Figure 3-5. The Power Module

Experimental Setup

23

The main power card

The main power card, which holds all the power electronic devices, like motor bridges, high voltage transistors, large electrolytic capacitors and a mechanical relay and powers the electrical DC-motors on the wheelchair, is shown in Figure 3-7. The functionalities of the different components on the card were investigated with the help of reverse engineering. From these results, simplified electrical schemes for the different main parts of the power card have been created. There are four main parts on the main power card; the motor drivers, the power control, the part that handles the braking of the motors and finally the external connection socket.

The motor drivers

The motor part contains two identical full-bridge (H-bridge) motor drivers. A full-bridge circuit makes it possible to drive a motor both in backward and forward direction in opposite to a half-bridge circuit, which is only able to drive a motor in one direction. The full-bridge circuit on the power card is driven by an H-bridge FET driver IC called HIP4082 [10]. A simplified electrical scheme of the motor part, with the H-bridge driver HIP4082, on the main power card is shown in Figure 3-8.

BHO

BLO

ALO

BHS

AHS

AHO

BHI

BLI

ALI

AHI

HIP4082

12V

GND

MOTOR (LOAD)

GND

24V (BRIDGE)

PWM

12V

DIS

GND

EN/DIS HIP

Figure 3-7. The main power card

Figure 3-8. A simplified electrical scheme of the motor-bridge driver HIP4082 that controls the motor’s rotational speed by changing the PWM signal’s duty cycle.


24

The HIP4082 acts as a PWM (Pulse Width Modulation) motor controller, which means that the rotation speed of the motor can be altered by changing the duty cycle of the square wave signal. Depending on the duty cycle of the received PWM signal, the HIP4082 opens and closes the four transistors in the H-bridge at different intervals. For example, when the PWM signal goes high the HIP4082 will close T1 and T3 and open T2 and T4, resulting in the current flowing through the motor (load) from left to right (forward direction), as shown to the left in Figure 3-9. When the signal goes low, T1 and T3 will be open and T2 and T4 closed, resulting in the motor rotating in the opposite direction (the current will go from right to left), which is shown to the right in the same figure.

M +

T1 T4

T2 T3

M +

T1 T4

T2 T3

If the duty cycle is half the length of the square wave period (50 percent duty cycle), the switching time will be equally long and since the switching frequency is high (e.g. 40 kHz) the current will be zero resulting in a non rotating motor, because of the inertia in the motor. If the duty cycle is more than 50 percent, the motor will rotate in forward direction, since the resulting switched current will be flowing from left to right and vice versa. If it is less than 50 percent the motor will rotate backwards. In other words the rotation speed of the motor will be proportional to the duty cycle. The HIP4082 has an input EN/DIS HIP that is used to enable/disable the reading of the PWM signal. When disabled, the HIP4082 closes all transistors and no current will flow through the bridge.

The power control

The power control part on the main power card has three power-up stages; the IC POWER-UP, the STORAGE POWER-UP and the RELAY ON. All three signals need to be enabled in the mentioned order, to be able to drive the electrical motors. A simplified electrical scheme of the power control part, with the three enable signals is shown in Figure 3-10. The first stage is the IC POWER-UP, which powers up all components supplied with 12 volts by opening a p-channel junction gate field-effect transistor (JFET), which in turn supplies a 12 volt regulator with 24 volt. Examples of 12 volt components are the bridge driver HIP4082 and the microprocessor on the external processor card. The next stage is the STORAGE POWER-UP, which opens a second JFET transistor that charges six 1000 nF electrolytic storage capacitors with 24 volts. The storage capacitors work as a power

Figure 3-9. Right: The current through the motor flows from left to right Left: The current flows in the opposite direction

Experimental Setup

25

backup close to the motor bridges and is used to avoid voltage drops, if the motors momentarily draw a high current. The six large electrolytic capacitors can be seen on the main power card in Figure 3-7. This stage also prepares the mechanical relay to be turned on by applying 24 volts to one side of the coil. The final stage is called the RELAY ON, which opens a third transistor. In this case it is a NPN bipolar junction transistor (BJT) that grounds the other side of the relay coil, which switches the relay on. When the relay is turned on, the storage capacitors and the two motor bridges are directly connected to the 24 volt battery. As a result of this, the battery will be able to deliver high currents to the electrical motors. To be able to verify that the relay is turned on, an output signal VERIFY RELAY ON is used. This signal goes low when the relay is turned on, otherwise the signal is high (24 volt). If the relay is not turned on when the motors are driven the current will flow through the both JFET transistors and the fuse. The high current will destroy the fuse, saving the transistors and the copper tracks. The three diodes to the left in the scheme are used for blocking the current from flowing in the wrong direction, when the motors work as generators. The diode next to the relay protects the BJT transistor from being destroyed by the high coil current, when the transistor switches the relay off.

G

S

D

24V (BAT.)

IC POWER-UP

IN OUT

GND

GND

12V REG

12V

S

G

D

STORAGE POWER-UP

FUSE

GND …

STORAGE CAPACITORS 6 x 1000 nF

24V

RE

LA

Y

E

B

C

24V (BAT.)

24V (BRIDGE)

GND

RELAY ON

VERIFY RELAY ON

Motor brake handling

The motor brake handling part on the main power card is not well investigated, since the brakes are not going to be used in this project. This motor brake handling controls the electromagnetic brakes, placed on top of the motors. There are two control signals for each motor brake. One of the signals is probably used to control the current through the electromagnetic brakes. The braking force is opposite proportional to the current applied. When no current is applied the braking force is maximized. The other signal is most likely used to verify that the electromagnetic brakes are correctly connected.

Figure 3-10. A simplified electrical scheme of the power control part on the main power card


26

External connection socket

On the main power card there is a 22-pole socket to connect the external processor card, shown in Figure 3-11. Via the socket’s connections it is possible for the microprocessor on the external card to control and monitor the different functions on the main power card. The control signals PWM and EN/DIS HIP for the motor driver part and the three power-up signals IC POWER-UP, STORAGE POWER-UP and RELAY ON are, among others, accessed through this socket. For more details about the socket’s different connections read Appendix C.

External Processor Card

The external processor card, shown in Figure 3-12, is connected to the main power card through the 22-pole socket. On the external card there is a microprocessor that controls the main power card. For example, the microprocessor generates PWM signals to the motor drivers, handles the different power-up stages and monitors different verification signals. The microprocessor also handles the serial communication with the joystick module’s internal microprocessor. The serial communication works as an ordinary UART-protocol. The data is transmitted asynchronously at a rate of 19200 bits/second. Data bytes are sent in frames consisting of one startbit, 8 data bits, one parity bit and one stopbit. The serial communication protocol uses different types of packets, for example the joystick packet, the true data packet and the startup packet. The joystick packet, which is sent from the joystick module to the power module, consists of 6 bytes data which, among other aspects, contains information relating to the joystick’s current position (speed and direction). The microprocessor on the external card interprets the joystick packet and sends control signals to the main power card. The microprocessor on the external card is also able to detect different errors. For example, if the motors or the electromagnetic brakes are unplugged, the microprocessor sends an error message to the joystick module, which in turn displays it for the user on the bar graph display. The controller in the microprocessor on the external card is very simple, only following the commands from the joystick.

Figure 3-11. The 22-pole socket on the main power card

Figure 3-12. The external processor card

Experimental Setup

27

Figure 3-13. The wheelchairs motor part with the different gearbox ratios

3.1.3 Electrical DC-Motor with Gearbox

There are two electrical DC-motors, one on each wheel, on the motorized wheelchair. The motor, shown in Figure 3-13, is connected to a gearbox, which increases the torque and decreases the rotational speed proportionally. On the output shaft of the first gearbox there is a cogwheel. This cogwheel is in contact with a second cogwheel, which can be decoupled from the big cogwheel. When decoupled, the big cogwheel, together with the wheelchair’s wheel, will spin freely. The DC-motors are manufactured by Exmek Electric [11]. The motor should be powered by 24 volts and is then capable of generating a maximum torque of 3 Newton meter and a rotational speed of 170 revolutions per minute after the gearbox (on the small cogwheel to the left in the figure). The gear ratio from the motor axis to the first cogwheel is 27:1. From the first cogwheel to the wheel, the gear ratio is 3:1. The maximum torque on the wheel will therefore be 9 Nm (3Nm x 3) with a rotational speed of 56.7 rpm (170 rpm / 3).

GGeeaarrbbooxx rraattiioo 2277::11

rraattiioo 33::11

rraattiioo 11::11

MMoottoorrbbrraakkee

DDeeccoouuppllee CCooggwwhheeeell

On top of each motor there is an electromagnetic (solenoid) brake mounted, which is shown in the figure. These brakes are controlled by the power module and are used to brake the motor axis when no current is applied to them.

3.1.4 Batteries

The batteries, which supplies the motorized wheelchair, are two 12 volts rechargeable batteries connected in series, where each battery is able to deliver 12 ampere-hours. The batteries last approximately 3 hours on each charge.


28

Figure 3-14. The controller card 3.0

3.2 Modifications

The original joystick controlled motorized wheelchair had to be modified to a disturbance observed power assisted wheelchair, where the speed is measured and a programmable controller is used to control the wheelchair.

3.2.1 New Control Hardware

The original motorized wheelchair is controlled with the help of a joystick module, described in Chapter 3.1.1, which communicates with the power module. This joystick module has now been removed, since the new disturbance controlled wheelchair is going to work as an ordinary wheelchair, where the arms are used to propel the wheelchair. Instead, a processor with an embedded control system will control the power module, depending on the rotational encoder’s speed values. To be able to program and develop this embedded control system, a controller development board with a programmable microprocessor and analog input and output signals had to be constructed. The result was the controller card 3.0. The controller card 3.0, shown in Figure 3-14, has been developed at Halmstad University by Jonas Johansson and Daniel Petersson in cooperation with Tommy Salomonsson.

The controller card 3.0 is a development board, which can be used to implement and build up different control systems. The card features many different useful functions, for example two incremental encoder counters, four analog to digital converters and one digital to analog converter. The converters are able to handle both positive and negative voltage inputs. In addition there are, among others, access to two pulse width modulated (PWM) signals, three interrupts, address and data bus, SPI interface and five different voltage levels. The development control board has the ability to communicate with, for example, Mathwork’s MATLAB via a RS232 interface (serial communication) trough a 9-pole D-Sub port. This communication can be used for receiving and transmitting data and variables between the control card and MATLAB.

Experimental Setup

29

Further details about the communication with MATLAB can be read in Chapter 3.2.6 and more details about the controller card 3.0’s different functions and components can be found in Appendix A.

3.2.2 Replace Joystick Module with Control Card

Since the controller card 3.0 will be used, instead of the joystick module, to control the wheelchair, a decision was made to connect the control card directly into the power module’s hardware in favour of connecting it to the power module’s serial communication port, which was used for the joystick. Using the serial communication would have involved implementing a similar serial protocol, which was used between the joystick module and the external processor card in the power module, mentioned in Chapter 3.1.2. If the 9-pole serial port on the controller card had been used to communicate with the power module, the port could not have been used to communicate with MATLAB, which is a very important function. Furthermore, problems may come up with the serial protocol. For example, since the motor brakes are decoupled, the external processor card would send error packets to the control card and would probably ignore the control commands resulting in a non-moving wheelchair. A decision to connect the control card directly into the power module’s hardware gave full control of the power module without any compromises. The control card is connected to the power module’s main power card via the 22-pole socket, shown in Figure 3-11. The signals that are going to be controlled via the socket are the four pulse width modulated motor control signals, the three power-up signals and the enable signal to the motor controller chip. These signals can not be directly connected to the controller card 3.0, so an external middle card had to be constructed.

3.2.3 External Middle Card

This card that is placed inside the power module, shown in Figure 3-15, works as a middle card (between the control card and the main power card) that translates the control signals before entering the main power card. This middle card also works as a safety card and protects the power module from being destroyed by careless coding. The two pulse width modulated motor control signals ALI and BLI, which control the motor speed, can be controlled by just one PWM signal from the controller card, by using a logic inverter, since BLI should be the same as ALI but inverted. The middle card has two inverters, one for each motor driver, which makes it possible to control both motors with only two PWM signals. The three power-up signals IC POWER-UP, STORAGE POWER-UP and RELAY ON, which is mentioned in Chapter 3.1.2, are used to power-up the different parts in the power module, before driving the motors. The handling of these signals and the signal EN/DIS HIP, which is used to enable or disable the motor controller, are crucial. It is

Figure 3-15. The external middle card


30

important not to enable the motor controller and start driving the motors, before the relay is turned on. To guarantee that this will not happen, two safety functions are implemented on the middle card. The first safety function makes it impossible to enable the motor controller, without having enabled the relay first. The second function protects the user from loosing control over the wheelchair, if the control card looses connection to the power module while driving. The second function also protects the power module from being destroyed, if the control card is turned off when the battery power is applied. The schematic design of the external middle card’s can be viewed in Appendix D.

3.2.4 Power Module Modifications

Some modifications to the power module’s housing have been made, to make the connections to the controller card good-looking and easy to decouple. The left outlet on the power module, which was used to connect the joystick module, has been replaced by a 7-pole female XLR connector. This new outlet C1 will be used for connecting the controller card 3.0. The control card gets its power supply and controls the main power card through the 7-pole connector. The right outlet, which was used to connect optional devices, has been replaced by a 5-pole female XLR connector. This outlet C2 can be used to measure the voltage on the positive and negative pole on both motors. Inside the power module, cables from the XLR connectors to the 22-pole socket and the external middle card are wired, which is shown in Figure 3-16.

Figure 3-16. The power module with replaced outlets and the external middle card connected to the main card via cables

Experimental Setup

31

3.2.5 Measure Rotational Speed

Two rotational encoders are going to be used, to measure the speed of the wheelchair. On the original joystick controlled wheelchair, electromagnetic brakes were mounted on top of each motor, described in Chapter 3.1.3. The electromagnetic brake is connected to the motor axis and brakes the motor, when no current is applied from the power module. These solenoid brakes were removed to be able to mount the rotational encoders, which measures the rotation of the motor axis, shown in Figure 3-17. The electromagnetic brakes are not necessary, since the motors can be used to brake the wheelchair. The two rotational encoders chosen for this project are Pepperl+Fuchs’ incremental encoder RVI58N-032K1R66N-00500 [12], which generates 500 pulses per revolution or 125 pulses in quadrature x1 mode setting, which will be used in this project. The incremental encoder has a 5 pole male XLR connector which is used to supply the encoder with 5 volts and receive the A and B channel’s pulse signals. The encoder’s metal housing has three mounting holes and the encoder’s shaft has a diameter of 6 mm. As calculated in Chapter 3.1.3, the gear ratio from the motor axis to the wheel is 81:1, resulting in the encoder generating 81·125 = 10125 pulses per wheel revolution and since the wheel’s diameter is 0.60 meter, the distance per pulse or pulses per meter will be:

][537060.0

10125][1086.1

1012560.0 4 erpulses/metorsemeters/pul =

⋅⋅=⋅ −

ππ

A mounting kit was constructed, to connect the incremental encoder with the motor shaft. The mounting kit, shown in Figure 3-18, consists of a round steel flange, a flexible shaft coupling, five M4 machine screws and twelve M4 nuts. The circular flange, manufactured at the steel workshop at the university, has five mounting holes; three for the encoder and two for the motor and a large hole, in the middle, for the motor axis. The flexible coupling is used to connect the 8 mm motor shaft to the encoder’s 6 mm shaft. With the help of the mounting kit and the rotational encoder, the speed measuring part of the wheelchair was assembled.

Figure 3-17. The mounted rotational encoder

Figure 3-18. The encoder mounting kit


32

3.2.6 Serial Communication with MATLAB

As mentioned in Chapter 3.2.1, the controller card has the ability to communicate with a computer using the program Mathwork’s MATLAB via the serial communication. This communication gives the possibility to change parameters in the control card and monitor real time values from the control card. This communication in MATLAB can be done by using the function “the good idea”, which has been developed at the university. “The good idea” gives the opportunity to stream up to four channels at once, in other words receiving four different variables like, for example, the velocity, the control signal or other useful variables. “The good idea” monitors and updates each variable (stream channel) in real time on a graph display, which is shown in Figure 3-19.

The function “the good idea” can also be used to set different variables in the control system implemented in the controller card. This makes it easy to change controller parameters during test drives of the power assisted wheelchair. Since the controller card uses a standard 9-pole serial communication port to communicate with the computer running MATLAB, the serial cable, which is connected between the wheelchair and the computer, can be replaced by, for example, two Bluetooth serial port plugs that makes the communication wireless.

Figure 3-19. MATLAB running the function “the good idea”, which gives the opportunity to study and save the streamed measuring data

Experiments and Results

33

4 Experiments and Results

In this chapter the experimental setup with the torque sensor free power assisted wheelchair will be used to implement different power assisting control system. The three different control designs; the control design by Hori and Oh presented in Chapter 2.2.2, the proportional control design and the proportional derivative control design proposed in Chapter 2.4.2, will be implemented in the control card for the wheelchair. Further details about the control card can be found in Chapter 3.2.1. The software code for the three controllers will be written in C language. The code is developed in Microchip’s MPLAB IDE [13], which is also used for programming and debugging the control card. The control card, which controls the wheelchair, sends data information to a computer running the function “the good idea” in MATLAB, mentioned in Chapter 3.2.6. The information can for example be the wheelchair’s velocity and the control signal to the motors. The data between the control card and the computer is transferred wirelessly with Free2move’s Bluetooth serial port plugs called F2M01 [14]. First, in this experiment chapter, the model of the power assisted wheelchair will be obtained to be able to simulate, calculate and implement the different control systems. After that, the three power assisting control designs will be implemented, tested and evaluated on the torque sensor free power assisted wheelchair.

4.1 Wheelchair Model

The wheelchair model can be obtained by investigating the wheelchair’s velocity response when applying a step signal to the motors, as mentioned in Chapter 2.2.1. The velocity of the wheels, which is measured with the rotational encoders, is recorded with “the good idea” in MATLAB by using the control card. The unit of the velocity is encoder pulses per second and the velocity 1 meter per second corresponds to 5370 pulses per second, which is calculated in Chapter 3.2.5. The motors’ speed is altered by changing the duty cycle of the square wave signals, from the control card, that are fed to the pulse width modulated motor controller as explained in Chapter 3.1.2. The duty cycle on the control card has a 10-bit resolution (0-1023), where the duty cycle 511 results in a non rotating motor and the values 1023 and 0 results in a maximum rotation in forward and backward direction respectively. The duty cycle will now be defined as a function of the control signal u.

1023_0,511_ <<+= cycledutyucycleduty The wheelchair’s velocity step response is first investigated in Chapter 4.1.1. The step response shows high velocity spikes in the beginning of the response, which are caused by play in the gearboxes. This problem with the play will be solved in Chapter 4.1.2. In Chapter 4.1.3, the model of the power assisted wheelchair process will be calculated. Finally in Chapter 4.1.4, an analysis of how the wheelchair model is affected by different user weights will be conducted.


34

4.1.1 Step Response

The step responses of the power assisted wheelchair when applying two step changes, in the control signal u, from 0 to 250 and back to 0, is shown in Figure 4-1.

0 1 2 3 4 5 6 7 8

0

50

100

150

200

250

Step

t (time) [s]

u co

ntro

ll [d

uty

cycl

e]

0 1 2 3 4 5 6 7 8

0

1000

2000

3000

4000

5000

6000Velocity (step response)

t (time) [s]

vr (

real

vel

ocity

) [p

ulse

s/s]

right

left

In the velocity figure, the blue (solid) line represents the velocity of the right motor and the red (dashed) line corresponds to the velocity of the left motor. The large spikes, in the beginning of the increasing velocity step response, are caused by play in the wheelchair’s gearboxes. Further details about the wheelchair’s gearbox design can be found in Chapter 3.1.3. The play causes the motor axis, on which the rotational encoder is mounted, to rotate freely, a couple of revolutions before the wheel starts to rotate. This play results in a very high acceleration of the encoder signal when the motor revs up without load. A high deceleration then follows when the motor starts to drive the wheel. This phenomenon can be seen in the acceleration step response of the right and left motor, which is shown in Figure 4-2.

0 1 2 3 4 5 6 7 8-1.5

-1

-0.5

0

0.5

1

1.5x 10

4 Acceleration (step response)

t (time) [s]

ar (

real

acc

eler

atio

n) [

puls

es/s

²]

right

left

This problem has to be solved in software, since the play in the gearbox can not be reduced.

Figure 4-1. Left: The two step signals (increasing and decreasing) with a changed control signal of 250 Right: The wheelchairs velocity step response

Figure 4-2. The acceleration response with large spikes


35

4.1.2 Play Reduction

The measurement spikes in the step response, which are caused by the play in the gearbox, can be reduced in different ways. One way is for example using some sort of Kalman filtering, which modifies the measured velocity data. Another approach is to use a Play passing method, which does not measure the velocity when the play problem occurs.

Kalman filtering

The principle of the Kalman filter is that it estimates the next value, in this case the velocity, and compares it with the measured value. If the estimated value differs too much from the measured value, the Kalman filter will rely more on the estimated value. The effect of the Kalman filter is shown in Figure 4-3.

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

500

1000

1500

2000

2500

3000

3500

4000

4500Velocity (original)

t (time) [s]

vr (

real

vel

ocity

) [p

ulse

s/s]

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

500

1000

1500

2000

2500

3000

3500

4000

4500Velocity with Kalman

t (time) [s]

vr (

filte

red

velo

city

) [p

ulse

s/s]

The filter damps the velocity spikes, which is a desirable effect. But, this filtering approach has however a disadvantage. The filter also damps the external disturbances caused by the user. Therefore a better solution, where the measurements are not modified, is needed.

Play passing method

Instead of changing the measured data to reduce the velocity spikes caused by the play problem, a play passing method that does not measure the velocity when the play problem occurs is used. The main idea of the method is that before the control system starts to measure the velocity it will apply a small force that is enough to rotate the motor axis past the play, but without start rotating the wheels. In other words, the motor axis together with the rotational encoder will not rotate freely on the start, which will reduce the large spikes in the beginning of the velocity response. For the play passing method to work in both forward and backward direction, it has a direction sensing function. This function tells the motors to rotate past the play in either forward or backward direction. Therefore, when the wheelchair user wants to change direction, this play reduction method will first close the power assisting control system, and then change direction of the play passing force and finally start the power assisting control system again.

Figure 4-3. Left: The original step response. Right: The Kalman filtered step response


36

As a result of this, the control system for the power assisted wheelchair will have two modes; the controller mode and the play passing mode, which the system jumps between. The state graph with the two modes is shown in Figure 4-4.

CCoonnttrroolllleerr mmooddee PPllaayy ppaassssiinngg mmooddee

0)1()( <−⋅ kykychange direction

treshold reached

0)( dcku th ⋅>

)(ku

Filter

))((0 kysignd ⋅)(ku)(ky

R

S−

Power assisting Controller

When the system is in the controller mode, the regular power assisting controller runs, which gives the user additional power. If the user changes direction, which results in a changed sign of the wheelchair’s velocity y(k), the system jumps to the play passing mode. In the play passing mode a small motor force is applied, which will rotate the motor axis together with the rotational encoder past the play, by changing the control signal u(k) to d0 ·sign(y(k)). To prevent this direction change from feeling jerky, the step signal d0 ·sign(y(k)) is run through a filter, which is shown in Figure 4-5. The response of the filter is altered by changing the time constant af, which should be chosen -1 < af <0.

u))((0 kysignd ⋅11

1−+

+qa

a

f

f

Filter

))((0 kysigndcth ⋅⋅

This filter slows down, or smoothes, the direction change. When the control signal u(k) reaches a certain threshold value cth· d0, the system will jump back to the controller mode. The threshold value can be adjusted with the threshold constant cth, which should be chosen 0 <cth < 1. When the system jumps back to the controller mode, the wheelchair’s power assisting controller is enabled again, and the user can drive in the other direction without the play in the gearbox disturbing the control system. The play reduction method works individually on each motor, which makes it possible for the user of the power assisted wheelchair to rotate the wheels in different direction, thus making a tight turn with the wheelchair. The only disadvantage with this play passing method is that it will cause a small delay, depending on the time constant af in the filter, before giving the assisting power when changing direction. The step response with and without the play reduction method is shown in Figure 4-6.

Figure 4-5. The filter that slows down the step signal, which is used to pass the gearbox play, and the jump back threshold

Figure 4-4. The state graph of the power assisted system with the play passing method


37

0 1 2 3 4 5 6 7 8

0

1000

2000

3000

4000

5000

6000

7000Velocity (original)

t (time) [s]

vr (

real

vel

ocity

) [p

ulse

s/s]

right

left

0 1 2 3 4 5 6 7 80

1000

2000

3000

4000

5000

6000

7000Velocity with the play reduction method

t (time) [s]

vr (

real

vel

ocity

) [p

ulse

s/s]

right

left

This play reduction method works well. The method reduces the spikes in the velocity response, especially in the increasing phase. However, the method will give a larger gain compared to the original step response with the same control signal change of 250. This larger gain is due to the extra play passing force with d0 = 45, which actually gives a control signal change from 45 to 295 and back to 45. From now on this play reduction method, with the parameter settings d0 = 45, af = -0.22 and cth = 0.98, will be used in all experiments with step responses and control systems.

4.1.3 Calculate the Wheelchair Model

The two parameters Jr and Br, mentioned in Chapter 2.2.1, which are used for the wheelchair model, were calculated by studying the velocity step response. The wheelchair’s velocity response when the duty cycle, used in the step, is increased with 250 and the user sitting in the wheelchair weighs 82 kg, is shown in Figure 4-7.

0 1 2 3 4 5 60

1000

2000

3000

4000

5000

6000

7000Velocity (step response, step increased 250, user weight 82 kg)

t (time) [s]

vr (

real

vel

ocity

) [p

ulse

s/s]

right

left

Since the velocity on the right and left motor are almost equal, the same model of the wheelchair will be used for both wheels.

Figure 4-6. Left: The original step response Right: The step response with the play reduction method

Figure 4-7. The wheelchair’s step response with a user weighing 82 kg


38

The inertia factor Jr and the damping factor Br, is then

==

⇒

===

===

0377.0

0397.0

95.0)3970()%63(

2.25250

63001

max

max

r

r

rr

r

inc

r

r

J

B

tvtB

J

u

v

B

The discrete-time process is then calculated by using the equations in Chapter 2.3.1. The number of samples per second is chosen to 20, which gives the sampling period h = 0.05 seconds. This choice of sampling period gives good parameter values in the discrete-time form, without the wheelchair feeling jerky. This choice also gives the control system more calculation time between each sample.

11

11

1

11

1)(

)()(

1)( −

−

−

−−

+==→

+=

qa

qb

qA

qBqP

BsJsP

rr

( ) ( ) 292.19487.010397.0

11

1

-0.9487

11

05.00377.0

0397.0

1

=−=+=

=−=−=⋅−−

aB

b

eea

r

hJ

B

r

r

The model of the real wheelchair process in discrete-time will then be:

1

1

1

11

949.0129.1

)()(

)( −

−

−

−−

−==

q

q

qA

qBqP

The model of the discrete-time process’ velocity step response is compared to the real process’ response in Figure 4-8.

0 1 2 3 4 5 60

1000

2000

3000

4000

5000

6000

7000Velocity (step response)

t (time) [s]

vr (

real

vel

ocity

) [p

ulse

s/s]

simulated

real

Figure 4-8. The wheelchair’s simulated velocity step response compared to the real velocity response


39

The simulated velocity step response follows the real measured velocity of the wheelchair very well. This means that the calculated wheelchair model agrees well with the real wheelchair process. But, there is one important difference between the simulated and the measured step response. The real velocity oscillates a bit, which has to do with the internal structure of the wheelchair that makes the wheels wobble. This oscillation will cause problems for the power assisting controllers, especially in the control design by Hori and Oh. More about this problem can be read in Chapter 4.2, when the controllers are implemented.

4.1.4 Weight Affects

How the weight of the user affect the wheelchair model is an important parameter to investigate. For example, one important question is, will the torque sensor free power assisted wheelchair still work if the user gains weight? If so, a function will be needed to adjust the wheelchair model depending on the weight. This investigation was carried out by looking at the velocity step response for different weights. The same duty cycle step change of 250 was used in all experiments. The different weights used were 21 kg, 57 kg, 82 kg and 103 kg. The results of the weight experiment are shown in Figure 4-9.

0 1 2 3 4 5 6 70

1000

2000

3000

4000

5000

6000

7000Velocity step response with different user weights (step increased 250)

t (time) [s]

vr (

real

vel

ocity

) [p

ulse

s/s]

weight 21 kg

weight 57 kg

weight 82 kg

weight 103 kg

This experiment shows that the final velocity is almost the same for all different user weights. But, the time response will become slightly slower for increasing weight. This means that the wheelchair model will not be significantly affected by small changes in the user’s weight. In other words, if the user gains weight during the day, the power assisted control system will still function in the same way. Therefore the calculated model in Chapter 4.1.3, with the user weight 82 kg, will be used when simulating and implementing the control systems.

Figure 4-9. The wheelchair’s velocity step response with different user weights


40

4.2 Controller Implementation

The three control designs, the control design by Hori and Oh presented in Chapter 2.2.2, the proportional control design and the proportional derivate control design proposed in Chapter 2.4.2, will be implemented on the power assisted wheelchair. First, in Chapter 4.2.1, the proportional control design will be implemented on the power assisted wheelchair. When implemented, this control design will give the user additional power, but the stability criterion limits the amplification of the user’s force. In Chapter 4.2.2, the Hori and Oh control design, which also gives the user assisting power, but without a maximum limitation of the amplification of the force, will be implemented. But this controller design is sensitive to modelling errors, which makes it difficult to make the design function properly in reality. However, the desired amplification of the user’s force can also be obtained by adding a derivative part to the proportional controller. This proportional derivative control design will be implemented in Chapter 4.2.3. The method when implementing each controller, when using the calculated model in Chapter 4.1.3, is divided into four parts. First, the stability limits for the control parameters, are computed. After that, the velocity response of the power assisted wheelchair, with and without the assisting controller, is simulated in MATLAB when applying a known disturbance. The same known disturbance is then applied to the implemented controller on the wheelchair’s controller card and the measured velocity is compared to the simulated values. Finally, the implemented control system on the power assisted wheelchair is evaluated.

4.2.1 Proportional Controller Design

The proportional control design, which is proposed in Chapter 2.4.2, is the simplest kind of controller that gives power assistance. This controller, shown in Figure 4-10, only contains one control parameter, which is the gain constant K, that should be chosen negative to give assisting power.

u yK−

The larger the negative value of K, the more assisting power the wheelchair will give the user. But, the proportional controller will have a maximum amplification limit of the force applied by the user.

Parameter limits

For the proportional controller to give additional power, the gain constant K has to be negative. The stability criterion for the closed loop system 11 <λ where ( )111 Kba +−=λ and that K < 0

gives the limitation:

01

1

1 <<−−K

b

a

Figure 4-10. The P-controller


41

With the wheelchair model values a1 = -0.949 and b1 = 1.29 the K limits will be:

00395.0 <<− K

Simulation

First the control system with the proportional controller, shown in Figure 4-11, will be simulated. The process model of the power assisted wheelchair, which is calculated in Chapter 4.1.3, is used in the simulation. When simulating, an external disturbance signal d corresponding to a user propulsion, is fed to the system. The signal is a square pulse with a top value of 200 and a length of one second. This pulse signal is a rough simplification of one propulsion movement.

d

+ y

1

1

949.01

29.1−

−

− q

q

+

K−

Wheelchair model

Proportional controller

The simulated response for the gain constant K = -0.030 when the control system is fed with the disturbance pulse signal is shown in Figure 4-12.

0 1 2 3 4 5 6 7 8 9 10

0

50

100

150

200

Disturbance pulse signal

d (d

istu

rban

ce)

[dut

y cy

cle]

t (time) [s]0 1 2 3 4 5 6 7 8 9 10

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000Simulated velocity (disturbance response)

y (v

eloc

ity)

[pul

ses/

s]

t (time) [s]

without controller

P-cont, K=-0.030

The simulation shows that the power assisted wheelchair with the proportional controller has a larger maximum velocity and a longer duration compared to the same wheelchair without any controller, when applying the disturbance pulse signal.

Figure 4-11. The power assisting control system with the proportional controller

Figure 4-12. Left: The disturbance signal that corresponds to a user propulsion Right: The disturbance response with and without the proportional controller


42

Implementation

After the simulation, the proportional controller is implemented in the controller card on the power assisted wheelchair. Then, the same disturbance pulse signal that was used in the simulation is fed to the wheelchair’s control system. The wheelchair’s disturbance response is measured and sent wirelessly to MATLAB. The measured velocity data is compared to the simulated data in Figure 4-13.

0 1 2 3 4 5 6 7 8 9 10

0

50

100

150

200


d (d

istu

rban

ce)

[dut

y cy

cle]

t (time) [s]0 1 2 3 4 5 6 7 8 9 10

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000 Velocity (disturbance response), P-controller K = -0.030

y (v

eloc

ity)

[pul

ses/

s]

t (time) [s]

real

simulated

The measured velocity data matches the simulated disturbance response well. The difference after six seconds has to do with a slightly uneven floor, which makes the wheelchair turn a little and vary in velocity. The two spikes, one in the beginning and one on the top, are caused by the play problem, which is described in Chapter 4.1.2.

Evaluation

The proportional controller, implemented on the power assisted wheelchair, works surprisingly well, despite its simplicity.

0 1 2 3 4 5 6 7 8 9 10

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000


y (v

eloc

ity)

[pul

ses/

s]

t (time) [s]

real

simulated

0 1 2 3 4 5 6 7 8 9 10

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000


y (v

eloc

ity)

[pul

ses/

s]

t (time) [s]

real

simulated

Figure 4-13. Left: The disturbance signal that corresponds to a user propulsion Right: The real and the simulated disturbance response

Figure 4-14. Left: The real and the simulated disturbance response with the gain constant K=-0.025 Right: The real and the simulated disturbance response with the gain constant K=-0.035


43

The controller amplifies the user’s propulsion and generates a longer lasting propulsion influence. But, the amplification of the initial force is limited. If the parameter K in the controller is increased, the amplification of the initial force is increased, but that will also make the influence response to last longer. The influence of the proportional controller with the gain values -0.025 and -0.035 is shown in Figure 4-14. If the gain constant K is raised above the stability limit, the power assisted wheelchair will accelerate to maximum velocity and be stuck there after an external disturbance.

4.2.2 Control Design by Hori and Oh

The power assisting control design by Hori and Oh, which is presented in Chapter 2.2.2, is a clever design, at least in theory. The main idea of this controller is that it obtains the external disturbance by the user, by using a process eliminating function and then feeds the disturbance to a desired process model of the wheelchair. This control design, shown in Figure 4-15, contains three control parameters; the time constant am and the gain factor bm, which are used in the desired wheelchair model, and the constant K, which is used to regulate the power assisting controller’s influence.

( )( )KBA

KbbAA

mm

mm

+−− 1

u y

or

u y ( ) ( )1

1111

)(1

)()(1−

−

−+−+−

−qaKb

qabbaKbbK

mm

mmm

Parameter limits

The three control parameters am, bm and K should be chosen within some limitations to give the desired effect. The desired gain factor bm should be chosen greater than the real wheelchair’s gain factor b1 to give a higher velocity on each push. The desired time constant am should be closer to -1 than the real time constant a1 to achieve a longer lasting response of each push. The constant K should be greater than zero. These limitations together with the stability criterions for the

closed loop system, which are 12,1 <λ where ma−=1λ and ( )112 Kba +−=λ , give the limitations:

1

111́

10,,1

b

aKbbaa mm

−<<<<<−

With the wheelchair model values a1 = -0.949 and b1 = 1.29 the three limits will be:

51.10,29.1,949.01 <<<−<<− Kba mm

Figure 4-15. The controller by Hori and Oh


44

Simulation

The controller by Hori and Oh, shown in Figure 4-16, will be simulated. The calculated process model of the power assisted wheelchair is used in the simulation. When simulating, an external disturbance signal d corresponding to a user propulsion, is fed to the system. The signal is a square pulse with a top value of 200 and a length of one second.

d

+

1

1

949.01

29.1−

−

− q

q

+ Wheelchair model

Controller by Hori and Oh

y

( ) ( )1

1111

)(1

)()(1−

−

−+−+−

−qaKb

qabbaKbbK

mm

mmm

The simulated disturbance response with the parameters am = -0.99, bm = 1.8 and K = 1 when the control system is fed with the disturbance pulse signal is shown in Figure 4-17.

0 1 2 3 4 5 6 7 8 9 10

0

50

100

150

200


d (d

istu

rban

ce)

[dut

y cy

cle]

t (time) [s]0 1 2 3 4 5 6 7 8 9 10

0

1000

2000

3000

4000

5000

6000


y (v

eloc

ity)

[pul

ses/

s]

t (time) [s]

without controller

H&O-cont, am=-0.99, bm=1.8 K=1

The simulation shows that the power assisted wheelchair with the Hori and Oh controller has a larger maximum velocity and a longer lasting response compared to the same wheelchair without any controller, when applying the disturbance pulse signal.

Figure 4-16. The power assisting control system with the Hori and Oh controller

Figure 4-17. Left: The disturbance signal that corresponds to a user propulsion Right: The disturbance response with and without the Hori and Oh controller.


45

Implementation

The control design by Hori and Oh is implemented in the controller card on the power assisted wheelchair. Then, the same disturbance pulse signal that was used in the simulation is fed to the wheelchair’s control system. The wheelchair’s disturbance response is measured and the measured data is compared to the simulated data in Figure 4-18.

0 1 2 3 4 5 6 7 8 9 10

0

50

100

150

200


d (d

istu

rban

ce)

[dut

y cy

cle]

t (time) [s]0 1 2 3 4 5 6 7 8 9 10

0

1000

2000

3000

4000

5000

6000

7000 Velocity (disturbance response), Hori & Oh-controller, am=-0.99, bm=1.8 K=1

y (v

eloc

ity)

[pul

ses/

s]

t (time) [s]

real

simulated

The power assisted wheelchair’s velocity does not match the simulated response at all. The controller generates a bad result and is not able to amplify the external disturbance in a desired way, despite using the play reduction method. This controller-malfunction is mainly caused by the play in the wheelchair’s gearbox, which is explained in Chapter 4.1.2, and the oscillation in the wheelchair’s velocity, which was noticed in the step response in Chapter 4.1.3. These problems make the velocity signal incoherent. Since this controller theoretically compares the present velocity sample with the previous velocity sample, in other words some sort of acceleration, the control signal will also be incoherent and terrible. The control signal and the acceleration, when applying the simulated disturbance, are shown in Figure 4-19.

0 1 2 3 4 5 6 7 8 9 10-600

-400

-200

0

200

400

600Control Signal

t (time) [s]

u (c

ontr

ol s

igna

l) [d

uty

cycl

e]

0 1 2 3 4 5 6 7 8 9 10-1.5

-1

-0.5

0

0.5

1

1.5x 10

4 Acceleration

t (time) [s]

ar (

real

acc

eler

atio

n) [

puls

es/s

²]

Figure 4-18. Left: The disturbance signal that corresponds to a user propulsion Right: The real and the simulated disturbance response with the controller by Hori and Oh

Figure 4-19. Left: The generated control signal with the controller by Hori and Oh Right: The acceleration disturbance response


46

The controller could have worked if the simple wheelchair model with only one time-constant, presented in Chapter 2.2.1, had been replaced by a more precise model, with the play and oscillation defects included.

Evaluation

The control design by Hori and Oh is not implementable in a practical point of view. Since the controller contains a process eliminating function, which is an inverse model of the real process, to obtain the external disturbance, the model has to be very precise. The wheelchair model that Hori and Oh have proposed differs too much from the real wheelchair process, which results in the controller not functioning properly on the wheelchair.

4.2.3 Proportional Derivative Controller Design

The proportional derivative control design, which is proposed in Chapter 2.4.2, contains besides the gain constant from the proportional controller also a derivative part that can be used to increase the users’ initial force even more, without changing the lasting time of the propulsion influence. The derivative part in the PD-controller is implemented as a high pass filter with a limited noise feedback, which reduces high frequency noise and incoherency in the velocity. This controller, shown in Figure 4-20, contains three parameters; the proportional gain Kp, the derivative gain Kd and the frequency limiter parameter N.

u y ( ) ( )1

2

1211

1 −

−

−−−++

−qc

qcKcKcKK pddp

where the constants c1 and c2 in the controller are:

NhK

Nc

d +=1 timesamplingh

NhK

Kc

d

d =+

=2

Parameter limits

For the proportional derivative controller to give additional power the gain constant Kp has to be negative as in the proportional controller. The outer stability limits of Kp will still be

01

1

1 <<−−pK

b

a

when Kd is set to zero, which gives an ordinary proportional controller. But the lower limit of Kp will increase with increasing Kd value. The stability limitations for three parameters Kp, Kd and N can be found in the area that is enclosed by 112 −−> gg , 112 −> gg and 12 <g where

( )111211 cKbKbcag dp ++−= ( )2111212 cKbcKbcag pd −−−=

Figure 4-20. The PD-controller


47

Simulation

The control system with the proportional derivative controller, which is shown in Figure 4-21, will be simulated. The process model of the power assisted wheelchair, which is calculated to a1 = -0.949 and b1 = 1.29 in Chapter 4.1.3, is used in the simulation. In the simulation, an external disturbance signal d corresponding to a user propulsion, is fed to the system. The signal is a square pulse with a top value of 200 and a length of one second.

d

+

1

1

949.01

29.1−

−

− q

q

+ Wheelchair model

Proportional derivative controller

y

( ) ( )1

2

1211

1 −

−

−

−−++−

qc

qcKcKcKK pddp

The simulated response for the gain constants Kp = -0.030 and Kd = -0.010 and with the frequency limiter parameter N = 0.02 when the control system is fed with the disturbance pulse signal is shown in Figure 4-22.

0 1 2 3 4 5 6 7 8 9 10

0

50

100

150

200


d (d

istu

rban

ce)

[dut

y cy

cle]

t (time) [s]0 1 2 3 4 5 6 7 8 9 10

0

1000

2000

3000

4000

5000


y (v

eloc

ity)

[pul

ses/

s]

t (time) [s]

without controller

P - cont, K=-0.030PD-cont, Kp=-0.030, Kd=-0.010, N=0.02

The simulation shows that the initial force is increased additionally with the proportional derivative controller compared to the proportional controller, without changing the lasting time so much, when applying the disturbance pulse signal.

Figure 4-21. The power assisting control system with the proportional derivative controller

Figure 4-22. Left: The disturbance signal that corresponds to a user propulsion. Right: The disturbance response without a controller, with P-controller and with PD-controller.


48

That the three chosen parameters Kp = -0.030, Kd = -0.010 and N = 0.02 fulfils the stability criterion, which is mentioned in parameter limits above, can be checked by calculating g1 and g2.

( )111211 cKbKbcag dp ++−= ( )2111212 cKbcKbcag pd −−−=

818.105.002.001.0

02.01 =

⋅+−=

+=

NhK

Nc

d

909.005.002.001.0

01.02 =

⋅+−=

+=

NhK

Kc

d

d

9213.0909.003.029.1818.1)01.0(29.1909.0)949.0(

9202.1818.1)01.0(29.103.029.1909.0949.0

2

1

=⋅−⋅−⋅−⋅−⋅−−=

−=⋅−⋅+−⋅+−−=

g

g

which fulfils the stability criteria

9202.019213.0 12 =−−>= gg , 9202.219213.0 12 −=−>= gg and 19213.02 <=g

Implementation

The proportional derivative controller is implemented in the controller card on the power assisted wheelchair. Then, the same disturbance pulse signal that was used in the simulation, is fed to the wheelchair’s control system. The wheelchair’s disturbance response is measured and the measured data is compared to the simulated data in Figure 4-23.

0 1 2 3 4 5 6 7 8 9 10

0

50

100

150

200


d (d

istu

rban

ce)

[dut

y cy

cle]

t (time) [s]0 1 2 3 4 5 6 7 8 9 10

0

1000

2000

3000

4000

5000

6000 Velocity (disturbance response), PD-controller Kp = -0.030 Kd = -0.010

y (v

eloc

ity)

[pul

ses/

s]

t (time) [s]

real

simulated

The measured velocity data matches the simulated disturbance response well. Once again (as in the proportional controller), the spikes at the beginning of the response, which is resulted by the play in the gearbox, and the difference after five seconds, which is caused by the slightly uneven floor, is shown in the wheelchair’s disturbance response.

Figure 4-23. Left: The disturbance signal that corresponds to a user propulsion Right: The real and the simulated disturbance response


49

The proportional derivative controller differentiates the velocity to get the acceleration, which is described in Chapter 2.4.2. But, since the velocity is discontinuous and noisy, the velocity is differentiated through a high pass filter with a limited noise feedback, which reduces the high- frequency gain in the acceleration. The filtered acceleration, with the frequency limiter parameter N = 0.02 and Kp = -0.030, compared to the real (unfiltered) acceleration is shown in Figure 4-24.

0 1 2 3 4 5 6 7 8 9 10-5000

0

5000

10000

15000

20000High frequency filtered acceleration with limited noise feedback

t (time) [s]

yd(a

) (a

ccel

erat

ion

) [p

ulse

s/s²

]

filtered, |Kd|/N = 0.05

real

The noise and the high spikes in the real acceleration are reduced greatly in the filtered acceleration, which gives a good result for the PD-control system. Otherwise, if the unfiltered acceleration had been used instead, the wheelchair would have felt very jerky when driving.

Evaluation

The proportional derivative controller amplifies the user’s propulsion and generates a longer lasting propulsion influence, which is a desirable effect.

0 1 2 3 4 5 6 7 8 9 10

0

1000

2000

3000

4000

5000


y (v

eloc

ity)

[pul

ses/

s]

t (time) [s]

real

simulated

0 1 2 3 4 5 6 7 8 9 10

0

1000

2000

3000

4000

5000


y (v

eloc

ity)

[pul

ses/

s]

t (time) [s]

real

simulated

Compared to the proportional controller, the PD-controller does not have the same limitation problem in the amplification of the initial force. The parameter Kp corresponds to K in the

Figure 4-24. The filtered differentiated velocity compared to the real acceleration

Figure 4-25. Left: The real and the simulated disturbance response with Kp=-0.030 and Kd=-0.005 Right: The real and the simulated disturbance response with Kp=-0.030 and Kd=-0.015


50

proportional controller and affects the disturbance response in the same way. If Kp is increased, the amplification of the initial force is increased, but that will also make the influence of the disturbance to last longer. However, Kd in the PD-controller can be used instead of Kp to increase the initial force without changing the lasting time. The influence of the proportional controller with the gain value Kp = -0.030 and the two different Kd values -0.005 and -0.015 is shown in Figure 4-25. As shown in the figure, when increasing Kd, the amplification of the initial force is increased.

Conclusions

51

5 Conclusions

This project resulted in a new kind of power assisting wheelchair, shown in Figure 5-1, that does not use torque sensors in the pushrims. This torque sensor free power assisted wheelchair only relies on one feedback signal. This feedback signal is the wheelchair’s velocity, which is measured with rotational encoders. The assisting power of the wheelchair can be adjusted in both how much the user’s force should be amplified and how long time the influence of the user’s propulsion should last. These adjustable parameters make it possible to configure the power assistance to suit all users, from those who just need a little additional assistance to those who need much assisting power. The torque sensor free power assisted wheelchair will behave as an ordinary wheelchair, but the user propelling the torque sensor free wheelchair will feel like his strength is increased sharply. As a result of this, it will be easier to propel the wheelchair in uphill direction, since the control system amplifies the user’s propelling movements. It will also become easier to brake the wheelchair in downhill direction since the control system amplifies the braking force applied by the user. A large part of the project was to build an experimental setup where an ordinary joystick controlled wheelchair had to be modified and rebuild. The power module, which drives the motors, has been modified and rotational encoders have been mounted on each motor axis. A controller card, which is used to implement and debug the power assisting control system, have been designed and constructed. This card also has the possibility to stream valuable data wirelessly via Bluetooth plugs to a computer, running the function “the good idea” in MATLAB. When the torque sensor free power assisted wheelchair was build, three different control designs where simulated, implemented and tested. The first design was a regular proportional control design, but instead of negative feedback this controller uses positive feedback, which amplifies the influence of the user’s force resulting in the wheelchair giving the user additional power. But, the stability criterion of the proportional controller gives a limitation of the amplification of the user’s force. The second design, implemented on the controller card, was a control design presented by Hori and Oh that also should give the user assisting power without the amplification limitation of the force, at least in a theoretical point of view. However, this design is very sensitive to wheelchair modelling errors, which makes it difficult to get this control design to function in a practical point of view. This has to do with the velocity signal being noisy and incoherent, mostly due to the play in the gearbox. The final design, which was implemented, was the proportional derivative controller, which has the possibility to amplify the force by using the derivative part, in other words the wheelchair’s acceleration. Since the velocity signal is noisy, the derivative part is implemented as a high pass filter with a limited noise feedback, which reduces the high-frequency gain. This proportional derivative controller gives the best torque

Figure 5-1. The new power assisted wheelchair


52

sensor free power assisted control system, which is robust and insensitive to measurement noise, where the amplification of the user’s force is adjusted with one parameter and the lasting time of the response with another parameter. This torque sensor free wheelchair has many advantages over the pushrim activated power assisted wheelchair. One big advantage is for example, if an assistant wants to push a functionally impaired user sitting in a power assisted wheelchair with torque sensors in the pushrims, the assistant will not benefit from the assisting control system. But, in the case with the disturbance observed controller system, which relies on the wheelchairs velocity, the wheelchair will also give additional power for the assistant. Another advantage is that the torque sensor free power assisted wheelchair has a simpler and cheaper structure with fewer components compared to the pushrim activated one. The torque sensor free wheelchair can have standard wheelchair wheels without a clumsy hub with torque sensors. The torque sensor free power assisted wheelchair can be improved further in different ways in the future. For example, the play in the gearboxes, which now is reduced in software, can instead be reduced in hardware by using a better gearbox with a lower gear ratio and fewer cogwheels, which is preferable. Another example is to install some sort of user panel that makes it possible for the user to adjust the assisting support for the two wheels individually in an easier way. This function can be used to make the wheelchair go straight even if the user has different levels of strength in his right and left arm. Furthermore, an improvement of the control system that considers and compares the velocities of both wheels can be developed in the future. This improvement can be used to make the wheels spin equally fast, thus making the wheelchair go straight despite unsynchronized propelling, in high velocities. Another function of the wheelchair that could be developed in the future is a weight sensing system that could tell the control system if there is a user sitting in the wheelchair or not. This weight sensing function could be used to turn on and off the control system automatically. Finally, a more detailed study of how the wheelchair behaves when driving in uphill and downhill direction could be performed in the future. This study could lead to a future function in the control system that adjusts the assisting power depending on the direction. This function could, for example, be used to help the user even more when driving in uphill direction.

References

53

References

[1] R. A. Cooper, T. A. Corfman, S. G. Fitzgerald, M. L. Boninger, D. M. Spaeth, W. Ammer

and J. Arva. Performane Assessment of a Pushrim-Activated Power-Assisted Wheelchair Control System. IEEE Transactions on Control Systems Technology, Vol. 10, Pages 121-126, January 2002.

[2] S. Oh and Y. Hori. Sensor Free Power Assisting Control Based on Velocity Control and

Disturbance Observer. IEEE ISIE 2005 - Dubrovnik Croatia, Pages 1709-1714, June 20-23 2005.

[3] H. Seki, T. Sugimoto and S. Tadakuma. Novel Straight Road Driving Control of Power

Assisted Wheelchair Based on Disturbance Estimation and Minimum Jerk Control. IEEE IAS 2005 – Hong Kong, Pages 1711-1717, October 2-6 2005.

[4] S. Oh and Y. Hori. Lateral Disturbance Rejection and One Hand Propulsion Control of a

Power Assisting Wheelchair. IEEE IECON 2005, Pages 1845-1850, November 6-10 2005. [5] H. Seki, T. Sugimoto and S. Tadakuma. Straight and Circular Road Driving Control of

Power Assisted Wheelchair Based on Balanced Assisted Torque. IEEE IECON 2005, Pages 451-456, November 6-10 2005.

[6] W. Li, N. Hata and Y. Hori. Prevention of Overturn of Power Assisted Wheelchair using

Novel Stability Condition. IEEE IECON 2005, Pages 1833-1838, November 6-10 2005. [7] Company. Decon Wheel. Hyltebruk Sweden, www.decon.se (December 2006) [8] Company. Be Rolka-Aktiv. Germany, www.berolka-aktiv.de (December 2006) [9] Company. PG Drives Technology. www.pgdt.com (December 2006) [10] Intersil. HIP4082 Data Sheet. January 3 2006. www.intersil.com/data/fn/fn3676.pdf

(December 2006) [11] Company. Exmek Electric. Shanghai China. www.exmek.com (December 2006) [12] Pepperl+Fuchs. Incremental Rotary Encoder RVI58N-*******6 Datasheet. August 1 2006.

www.pepperl-fuchs.com/selector/navi/productInfo/edb/t10721_eng.pdf (January 2007) [13] Company. Microchip. www.microchip.com (January 2007) [14] Free2Move. Bluetooth Serial Port Plug – F2M01C1 Datasheet. August 25 2004.

www.free2move.se/pdf/Datasheet_F2M01C1.pdf (January 2007)


54

Appendix A - Controller Card 3.0

55

Figure A-1. The controller card 3.0


The controller card 3.0 is a development board that can be used to implement and build up different control systems. The control card features many different useful functions. For example two incremental encoder counters, four analog to digital converters and one digital to analog converter, that all are able to handle both positive and negative voltage inputs. In addition there are, among others, access to two pulse width modulated (PWM) signals, three interrupts, address and data bus, SPI interface and five different voltage levels. The development control board has the ability to communicate with, for example, Mathwork’s MATLAB via a RS232 interface (serial communication). This communication can be used for receiving and transmitting data and variables between the control card and MATLAB.

The controller card 3.0, shown in Figure A-1, has been developed at Halmstad University by Jonas Johansson and Daniel Petersson in cooperation with Tommy Salomonsson. The card has been designed and routed with Orcad Capture CIS and Orcad Layout. The card was routed in two layers and the PCB itself was manufactured in 18 copies by Elprint. The card was assembled and soldered at the University. This development controller card will be used for the wheelchair project. Furthermore, the card will also be used for educational purposes at the University. The are six main parts on the control card; the microprocessor, the incremental encoder counter, the digital to analog and analog to digital converters, the RS232 communication, the external connection part and the part that handles the power supply for all components.


56

Figure A-2. The schematic design of the microprocessor part on the controller card 3.0

A.1 Microprocessor

The microprocessor on the control card is the PIC processor named PIC18F4525 manufactured by Microchip Technology. The PIC18F4525 is a 16-bit processor, which is clocked by a 40 MHz crystal oscillator. The PIC package is a 40 pins PDIP (Plastic Dual-In-Line Package) where 36 of the pins are used for I/O (Input and Output). The different input and output pins can be set into different configurations, giving the user for example up to 13 10-bit analog to digital converter inputs, 5 pulse width modulated signals, 8-bit data bus, 8-bit Address bus or 3 external interrupts.

The PIC18F4525 uses ICSP (In-Circuit Serial Programming) for programming and debugging code. Usually it is not possible to debug and step code in real time on a PIC processor, but with ICSP technology it is possible. In Figure A-2, the schematic design of the PIC processor on the control card is shown. To the left of the PIC18F4525 is the 40 MHz crystal oscillator, which clocks the processor. Above is the ICSP connection with a RJ12 6 pole modular outlet, which is used to connect a debugger/programmer. The push button is used to reset (restart) the flashed program, when the processor runs in stand-alone mode. The led to the right is used as an indicator that can be lit by setting the output bit B3 high.

R37 10k

R27 10k

Digital Out

Encoder CS#C100.1uF

VCC (5V)

AN_0AN_1AN_2AN_3

A0

C150.1uF TX

RXFCLK

A1

SDO

Y1

40 MHz

VCC4

N/C1

GND2

OUT3

VCC (5V)

SDA

R44330

CCP2

SW5

SW PUSHBUTTON

CCP1

D4LED

U13

PIC18F4525

MCLR/VPP/RE31

RA0/AN02

RA1/AN13

RA2/AN2/VREF-/CVREF4

RA3/AN3/VREF+5

RA4/T0CKI/C1OUT6

RA5/AN4/SS/HLVDIN/C2OUT7

RE0/RD/AN58

RE1/WR/AN69

RE2/CS/AN710

VDD111

VSS112

OSC1/CLKI/RA713

OSC2/CLKO/RA614

RC0/T1OSO/T13CKI15

RC1/T1OSI/CCP216

RC2/CCP1/P1A17

RC3/SCK/SCL18

RD0/PSP019

RD1/PSP120

RD2/PSP221RD3/PSP322RC4/SDI/SDA23RC5/SDO24RC6/TX/CK25RC7/RX/DT26RD4/PSP427RD5/PSP5/P1B28RD6/PSP6/P1C29RD7/PSP7/P1D30VSS31VDD32RB0/INT0/FLT0/AN1233RB1/INT1/AN1034RB2/INT2/AN835RB3/AN9/CCP236RB4/KBI0/AN1137RB5/KBI1/PGM38RB6/KBI2/PGC39RB7/KBI3/PGD40

SCL

C18100nF

VCC (5V)

C190.1uF

D7D6D5

VCC (5V)

D4

D3D2

R9110k

R92

470

D1D0

MCLR

DAC CS#

VCC (5V)

INT2INT1INT0

R23 10k

U17

ICSP

MC

LR1

VD

D2

VS

S3

PG

D4

PG

C5

R#

PGCPGD

Vref (2.5V)

W#R38 10k

A.2 Incremental Encoder Counter

The incremental encoder counter part, on the control card, can be used to count pulses from up to two rotational encoders.


57

Figure A-3. The schematic design of the incremental encoder part, with the counter LS266R1

A rotational encoder can be used to measure an axle’s angular position or rotation very precisely. Inside the encoder there is usually a coded disc made of metal, glass or plastic. On a photoelectric encoder the disc have a pattern of holes. A photoelectric encoder, for example, contains a LED and a light sensor and between them a disc with a pattern of hole. Each time the light from the LED passes through a hole to the light sensor, the encoder will generate a high signal. When there is no hole, the light will be blocked and the encoder will produce a low signal. As a result of this the encoder will generate a pulse (square wave) signal when the disc rotates. The rotational encoder’s resolution depends on the number of holes on the disc (more holes gives better resolution). Usually rotational encoders have a resolution ranging from 1 to 10,000 pulses per revolution. To be able to decide the direction of the rotation, most rotational encoders have two output pulse signals, called channel A and channel B. The B channel’s pulse signal is phase shifted 90 degrees with respect to channel A. By watching which channel’s pulse rises first, the direction of the rotation can be decided. For example, if channel A rises before B, the rotation direction is clockwise. In Figure A-3, the schematic design of the incremental encoder counter part is shown. The rotational encoders can be connected to the card through two 5 pole female XLR connectors (encoder X and encoder Y). Each of the connectors contains four signals; two output and two input signals. The two output signals are ground and 5 volt, which powers the connected encoder. The two input signals are the A and B channel’s pulse signals, which is generated by the encoder. The 24-bit dual axis counter LS7266R1, to the left in the figure, is used to count the pulses generated by the encoders. The LS7266R1 counts the pulses independently from the PIC processor. Therefore, the counter can be seen as a separate processor, which is clocked by the PIC processors FCLK signal instead of a crystal oscillator. This gives the possibility to read the counter value asynchronous (whenever the PIC processor wants) without using interrupts. Since the counter is a 24-bit system each encoder channel (X and Y) is able to count up to 16777215 pulses before it starts from 0 again. The LS7266R1 has an ordinary control interface with a 8-bit data bus, 2-bit address bus and the three signals; chip select, write enable and read enable. The 2-bit address bus gives the possibility to read the whole 24-bit value in three 8-bit parts through the data bus. With the control interface it is also possible to configure the counter chip into different modes and settings.

U18

LS7266R1

LCY1

Fclk2

5V3

D04

D15

D26

D37

D48

D59

D610

D711

GND12

A013

W#14

CS#15R#16A117ECX18LCX19Ch AX20Ch BX21XFLG122XFLG223Ch BY24Ch AY25YFLG226YFLG127ECY28

A0

JP2

Encoder X12345

Encoder CS#

D7D6D5D4D3D2

C140.1uF

D1

W#

D0

R#

FCLK

VCC (5V)

VCC (5V)

A1

VCC (5V)VCC (5V)

5-pole XLR

5-pole XLR

JP3

Encoder Y12345


58

Figure A-4. The schematic design of the analog to digital conversion part

A.3 Analog to Digital Conversion

The analog to digital conversion part, on the control card, can convert up to four analog signals into 10-bit digital values. The input signals to the A/D converter can be in different voltage ranges. There are two voltage modes on the converter part. The first mode can translate voltage levels between 0 and 5 volt or higher. The voltage range can for example be 0 to 5 or 0 to 10 volt. The second mode translates signals, where the voltage level is centred on 0 and differs 5 volt or more in both directions. The range can for example be -5 to +5 or -10 to +10 volt. It is also possible to apply low-pass filtering of the analog input signal, before it is digitalized by the PIC-processor. The low-pass filter is a first order filter, which removes high frequency noise in the analog signal. In Figure A-4, the schematic design of the analog to digital conversion part on the control card is shown. The A/D converter part contains four identical converters with two operational amplifiers each.

-

+

U24B

OP484

5

67

R43 100k

R1 100k -

+

U24C

OP484

10

98

R35 100k

R8 100k

VCC (5V)

-

+

U24D

OP484

12

1314

11

R46 100k

R15 100k

-

+

U24A

OP484

3

21

4

-

+

U23A

OP484

3

21

4

-

+

U23B

OP484

5

67

R2049.9k

R4549.9k

AN_0

AN_1

-

+

U23C

OP484

10

98

AN_3

AN_2

-

+

U23D

OP484

12

1314

11

A0_IN

R21 100k

A1_IN

VCC (-12V)

R3649.9k

A3_IN

A2_IN

A0

VCC (12V)

R22 100k

A1

A3

A2

Voltage Follower

R4749.9k

Low-Pass FilterNon-Inverting

Amplifier x3

R29 10k

R30 10k

R31 10k

R32 10k

R26100k

R3 100k

R4 100kR5100k

R10 100k

R11 100k R12100k

Vref (2.5V)

R17 100k

R18 100kR19100k

SW2Of f set Select

13

2J1

Resist

81

72

63

54

R25 100k

C3CAP NP

R24 100k

C4CAP NP

C5CAP NP

C6CAP NP

The analog-in signal AX_IN, to the left in the figure, can be divided down with the help of a selectable resistor. For example, if the voltage range is between 0 and 10 volt it needs to be divided down to a range between 0 and 5 volt since the A/D converter in the PIC-processor is limited to 5 volt. This is done with the help of a simple voltage division, by placing (selecting) a 100k ohm resistance in the J1 socket.


59

A0_IN

Analog OutDigital Out

A1_IN

A2_IN

A3_IN

P2

CONNECTOR DB15

8157

146

135

124

113

10291

Figure A-5. 15-pole D-Sub

Figure A-6. The schematic design of the digital to analog conversion part

Then the divided signal passes through a voltage follower. After the voltage follower there is an offset switch SW2, which is used to select between the two different voltage modes. If SW2 is switched to ground, the analog signal will be divided by three as a result of voltage division. If it is switched to 2.5 volt, this voltage will be added to the analog signal. This is used when the analog-in signal is both positive and negative for example -5 to 5 volt. This signal should first be divided with 2 before the voltage follower, resulting in a range from -2.5 to 2.5 volt. By adding 2.5 volt to the signal, the voltage range will be raised to be only positive, in other words a range from 0 to 5 volt. This signal will also be divided by three, before it is amplified three times by the non-inverting amplifier to the right in the figure. Finally, the analog signal passes through a low-pass filter, where the capacitor can be changed, to alter the cutoff frequency, or removed to keep the original signal. The modified analog signal is now ready to be digitalized by the PIC-processor. The four analog to digital inputs can be accessed through a 15-pole D-Sub connector, shown in Figure A-5.

A.4 Digital to Analog Conversion

The digital to analog conversion part, on the control card, can convert a digital signal from the PIC-processor into an analog output signal. The voltage range of the analog ouput signal can be anything between -10 and 10 volt. By changing software in the processor, the range can for example be -5 to 5, 0 to 5 or 0 to 10 volt. In Figure A-6, the schematic design of the digital to analog conversion part is shown. The D/A converter part contains a 12-Bit DAC chip called MAX530, which can deliver an analog signal between 0 and 4.096 volt. The chip is controlled by the PIC-processor with the help of a standard control interface with an 8-bit data bus and the three signals; chip select, write and read enable. The analog output signal VOUT from the MAX530 has a resolution of 1 millivolt per bit. The analog signal is then amplified 1.22 times with the help of a non inverting amplifier, resulting in a signal between 0 and 5 volt. The amplification can be fine-tuned to exactly 5 volt by a variable resistance. Finally the signal passes a differential amplifier, which lowers the range to -2.5 to 2.5 volt and then multiplies it with four, resulting in an analog output signal between -10 and 10 volt.

Amplifier x1.22

Non-Inverting

Amplifier (V-2.5)x4

Differential

A0 Analog Out

R392.1k

R4010k

A1R41 10k

R4240.2k

DAC CS#

VCC (5V)

W#

Vref (2.5V)

R14 10k

R16 40.2k

U21

MAX530

D1/D91

D2/D102

D3/D113

D44

D55

D66

D77

A08

A19

WR#10

CS#11

DGND12

REFIN13AGND14CLR#15LDAC#16REFGND17REFOUT18VSS19VOUT20RFB21ROFS22VDD23D0/D824

VCC (12V) VCC (12V)

VCC (-12V)VCC (-12V)

D2

C2033uF

Analog Out = 4 x (1.22 x VOUT - 2.5)

D0D1

D3

C97

0.1uF

-

+

U25A

OP484

3

21

411

-

+

U25B

OP484

5

67

411

D4D5D6D7

R28RESISTOR VAR (220)


60

Figure A-7. The serial communication part

Figure A-8. External Connections

The digital to analog output signal can be accessed through the 15-pole D-Sub connector, shown in Figure A-5.

A.5 RS232 Serial Communication

The RS232 serial communication part, on the control card, can be used to communicate with external devices, such as a computer, via a 9 pole D-Sub connector. The schematic design of the serial communication part is shown in Figure A-7. The driver/receiver chip, which is used to translate the different voltage levels in the communication, is Maxim’s MAX233A. The MAX233A is powered by 5 volts and this version uses internal capacitors, which eliminates the need of up to four external capacitors. The transmit signal Tx from the PIC-processor is translated to the correct voltage levels by the MAX233A chip and is transmitted out through the D-Sub. The signal received through the D-Sub is translated by the chip and this translated signal Rx is then received by the processor.

C160.1uF

U20

MAX233A

T2IN1

T1IN2

R1OUT3

R1IN4

T1OUT5

GND16

VCC7

(V+) C1+8

GND9

(V-) CS-10

C2+ (C2-)11V- (C2+)12C1- (C1+)13V+ (C1-)14C2+15C2-16V-17T2OUT18R2IN19R2OUT20

TXRX

VCC (5V)

P1

D-SUB

594837261

A.6 External Connections

On the control card there is a socket, which can be used to connect external devices. On the 26 pole socket, which is shown in Figure A-8, there is access to the following signals:

• Five different voltage levels; GND, 2.5 V, 5 V, 12 V and-12 V. • 8-bit data bus • 2-bit address bus • Read enable • Write enable • CCP1/PWM1 • CCP2/PWM2 • SPI interface with SCL, SDA and SDO • 3 external interrupts • FCLK

INT2INT1

VCC (5V)VCC (12V)

Vref (2.5V)VCC (-12V)

J2

X-Con

123456789

1011121314151617181920212223242526

INT0FCLK

R#

A0

W#CCP1

D3

D1D0

D5D4

D2

CCP2

D7D6

SCLSDA

A1

SDO


61

Figure A-9. The schematic design of the power supply part

A.7 Power

The power supply part on the control card feeds all the components with the correct voltage levels. The schematic design of the power supply part is shown in Figure A-9. This part contains two low dropout voltage regulators; one fixed 5 volt version called LM1084, which supplies the PIC-processor, the encoder counter chip, the RS232 chip and the DAC chip and an adjustable version, called LM1085, which is set to deliver 2.5 volt. The 2.5 volts are used to offset the voltage levels in the D/A and the A/D converters. Two LED:s are used to indicate that the regulators work. There is also a DC/DC converter on the control card. This is used to generate two new voltage levels. The converter is fed with 5 volt and it generates both positive and negative 12 volt. The ±12 volt is used as power supply to the operational amplifiers. To the left in the figure there is an outlet that is used to connect the power supply, which should be at least 7 volt. Next to the outlet there is a power switch, which is used to turn on and off the control card. The diode 1N5401 after the switch is used to prevent the card being damaged by wrong polarity, when connecting the power supply to the card.

VCC POWER

R710k

R1310k

C_DC/DC0.1uF

U12

LM1084

Out2

Gnd

1

In3

VCC POWER

C710uF

C810uF

D51N5401

C170.1uF

R33330

D1LED

DC/DC Conv erter +-12

X-Conn

+Vin1

-Vin (GND)2

-Vout4

Common5

+Vout6

R34330

D2LED

JP1

Power

123

C9610uF

VCC POWER

U11

LM1085

OUT2

AD

J1

IN3

VCC (5V)

C210uF

VCC (12V)

VCC (-12V)

SW Power

1

32

VCC (5V)Vref (2.5V)


62

Appendix B - Block Scheme Simplification of Hori and Oh’s Controller

63

Appendix B - Block Scheme Simplification of Hori and Oh’s Controller

The block scheme reduction procedure of the control design by Hori and Oh is shown below.

)(

)(1

1

−

−

qA

qB

d

u

K

1

1)(

b

qA −

1−q

)( 1−qA

b

m

m

rVy =

where x is introduced

A

B

d

u

K

11

−⋅m

m

A

b

b

A

1−q

m

m

A

b

y 0=r

K

x

where

( )

xKBA

Au

xAuKBA

uA

BKxuq

A

bKxu

mm

m

mmm

m

m

m

m

+=

=+

−=−= −1


64

which gives

A

B

y 0=r x

d

mm

m

KBA

A

+

m

mm

Ab

AbAbK

1

1−

where

( )( )

( )( )

( )( ) y

KBA

KbbAAu

yKBA

AbbAKu

yKBAb

AbAbKu

yAb

AbAbK

KBA

Au

mm

mm

mm

mm

mm

mm

m

mm

mm

m

+−−=

+−=

+−=

−⋅+

=

)(

)(

1

1

1

1

1

1

The block scheme simplification resulted in:

d

y

A

B 0=r u

( )( )KBA

KbbAA

mm

mm

+−− 1

Appendix C - Description of 22-pole Socket

65

Appendix C - Description of 22-pole Socket

On the power card there is a 22-pole socket, shown in Figure C-1, to connect the external processor card. A short description of the 22-pole socket’s different connections are described in Table 1.

Conn Name I/O Function 1 BLI M1 I Is used to control the rotational speed of motor 1 with the help of a PWM signal. 2 ALI M1 I Same as BLI M1 but with inverted PWM signal. 3 STORAGE

POWER-UP I Enables the charging of the storage capacitors. (Only works if the IC POWER-UP

is enabled) 4 VERIFY RELAY ON O Verifies that the relay have been turned on. 5 RELAY ON I Turns on the relay. (Only works if the STORAGE POWER-UP is enabled) 6 GND O Ground 7 12 V O 12 volts supply 8 IC POWER-UP I Turns on all 12 volt components. 9 24 V O 24 volts signal 10 BLI M2 I Is used to control the rotational speed of motor 2 with the help of a PWM signal. 11 ALI M2 I Same as BLI M2 but with inverted PWM signal. 12 GND O Ground 13 BRAKE M1 - 14 BRAKE M1 -

Probably used to control the current through the brakes and verify that the they are correctly connected

15 M1 + O Positive pole on motor 1 16 M1 - O Negative pole on motor 1 17 EN/DIS HIP I Enables the motor driver (Should only be enabled when RELAY ON is enabled) 18 M2 + O Positive pole on motor 2 19 M2 - O Negative pole on motor 2 20 BRAKE M2 - 21 BRAKE M2 -

Probably used to control the current through the brakes and verify that the they are correctly connected

22 GND O Ground

BLI

M1

MAIN POWER CARD

ALI

M1

STO

RA

GE

PO

WE

-UP

VER

IFY

RE

LAY

ON

RELA

Y O

N

GND

12 V

IC P

OW

ER

_UP

24 V

BLI M

2

ALI M

2

GN

D

B

RA

KE

M1

BR

AK

E M

1

M

1 +

M

1 -

E

N/D

IS H

IP

M

2+

M

2 -

B

RA

KE

M2

BR

AK

E M

2

GN

D

1 11

12 22

Figure C-1. Signal placement on the main power card’s socket

Table 1. Description of the 22-pole socket’s connections


66

Appendix D - Schematic Design of the External Middle Card

67

Appendix D - Schematic Design of the External Middle Card

The schematic design of the external middle card is shown in Figure D-1.

J1

8 HEADER

12345678

J2

HEADER 10

123456789

10

D115V

12

12VA ALIA BLI

B BLIB ALIRELAY#HIPGATE24V_INGNDGND

EN_GATEEN_RELAYEN_HIPB PWMA PWM5V

D25.15V

12

12V

LEFT MOTOR

24V_IN

RIGHT MOTOR

R16k

R21k

17.5V5V

U1A

74LS00

1

23

147

U1D

74LS00

12

1311

5V

A BLIA PWM

A ALI

B ALI

B PWM B BLI

U1C

74LS00

9

108

R610k

R5

2k

EN_GATEQ1BC549C

Q2BC549C

R7

2k

R833k

24V_IN

GATE

24V_IN

R933k

17.5V

U1B

74LS00

4

56

EN_RELAY

R310k

EN_HIP

R410k

#HIP

RELAY

Figure D-1. The schematic design of the external middle card

torque sensor free power assisted wheelchair …237836/fulltext01.pdfthe project resulted in a...

Documents