introduction to embedded systems & control: what...
TRANSCRIPT
Technische Universität München Technische Universität München
Introduction to Embedded Systems & Control: What are Cyber-Physical Systems?
Samarjit Chakraborty www.rcs.ei.tum.de
TU Munich, Germany
Technische Universität München
What are Cyber-Physical Systems?
Networked Embedded Systems Sensor network based systems Internet of Things (IoT) Embedded Control Systems
… will use automotive as a running example
2
Technische Universität München
This image cannot currently be displayed.
1993
2009
1 3 13
90 100
175
240
5 50
100
0
50
100
150
200
250
300
350
400
Memory [MB] ECUs
Automotive E/E Architecture Trends
1993
2009
3
Technische Universität München
The Modern Car
50 – 100 Electronic Control Units (ECUs) Complex in-vehicle network (CAN, FlexRay, Ethernet) Complex systems software, operating systems Several distributed control applications (safety-critical & comfort)
4
source: Bosch
Technische Universität München
5
Actuator System Sensor
Controller Network Network
τnetwork τnetwork
ACC ECU
Radar
Speed adjustment
Control Applications
Many of the safety-critical, driver assistance and comfort applications implement control functionality
Technische Universität München
Current Design Flow
6
High-level requirements
Applications
HW/SW architecture Multiple processor units (PUs) connected via a shared bus
… algorithm
design
Model
Task partitioning
T1 T2
T3
T4
Task mapping
m2 m1
Task and Message
scheduling
code generation
Algorithm & Architecture designed separately followed by integration, testing, debugging, …
Technische Universität München
Control Systems Design
7
System identification
Controller design
Control system analysis
system model controller
Equations
Technische Universität München
Control Systems Implementation
8
System identification
Controller design
Control system analysis
system model controller
Code generation
Task partitioning
Task mapping & scheduling
Message scheduling
Timing & performance analysis
Are control objectives satisfied
NO
Equations
Software
Technische Universität München
The Design Flow
9
System identification
Controller design
Control system analysis
system model controller
Code generation
Task partitioning
Task mapping & scheduling
Message scheduling
Timing & performance analysis
Are control objectives satisfied
NO
Controller Design
Controller Implementation
Technische Universität München
The Design Flow
10
Controller Design
Controller Implementation
Control theorist
Embedded systems engineer
Design assumptions Computing control law takes negligible time No delay from sensor to controller No delay from controller to actuator No jitter …
Implementation reality Tasks have non-negligible execution times Often large message delays Time and event-triggered communication
Technische Universität München
The Design Flow
11
Controller Design
Controller Implementation
Control theorist
Embedded systems engineer
These are implementation details
Model-level assumptions are not satisfied by
implementation
Not my problem!
Technische Universität München
Semantic Gap
12
Controller Design
Controller Implementation
Semantic gap between
model and implementation
Research Questions? How should we quantify this gap? How should we close this gap? Solution: Controller/Architecture Co-design
Technische Universität München
Cyber-Physical Systems
Tight interaction between computational (cyber) and physical systems
Isolated design of the two systems leads to: Higher integration, testing and debugging costs Poor resource utilization (e.g., unnecessarily high sampling
rates)
Question: Can existing techniques from real-time & embedded systems be
directly applied? We know hardware/software co-design
Answer: New techniques are required Traditional embedded systems design have focused only on
meeting deadlines Here, deadlines take a back seat
13
Technische Universität München
Dynamical Systems: DC Motor
14
The time dependence of the various states of dynamical systems is described by a set of difference or differential equations.
i
J = moment of inertia of the rotor b = system damping ratio, K = EMF and Torque constant, R = armature resistance, L = armature inductance, i = armature current, V = input, DC terminal voltage θ = output, position of shaft
Technische Universität München
Modeling Approaches: State-space
System states x(t): (1) motor shaft position, (2) motor shaft velocity, and (3) armature current, i
System Input u(t): Terminal voltage, V One might want to regulate one or more system states.
The regulated states are known as output y(t). E.g., motor speed or position
15
Technische Universität München
Performance Metrics
Transient Behavior:
16
The rise time tr is the time it takes the system to reach the vicinity of its set point or reference command The settling time ts is the time it takes the system transients to decay The overshoot Mp is the maximum amount the system overshoots (and often expressed as a percentage of reference command)
1
0.1
0.9
tr ts
Mp
t
Technische Universität München
Performance Metrics
17
Steady-state Behavior: Integral Input Cost:
- indicator of how much effort is needed for the control
Integral Error Cost: - indicator of how accurate the control performance is
There are many possible metrics. Moreover, one can also consider a weighted combination of the various cost such metrics
Technische Universität München 18
DC motor:
Objective:
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50
10
20
30
40
50
60
70
Seconds
Spee
d
33% Drop25% Drop0% Drop
A fraction of feedback signals being dropped
Controllers can be designed to be robust to
drops and deadline-misses
The deadlines are usually not hard for control-related messages
Control Tasks - Characteristics
Technische Universität München 19
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50
5
10
15
20
25
30
35
40
45
50
Seconds
Spee
d
Disturbance
h =20ms h =200ms
(1) The computation requirement at the
steady state is less, i.e., sampling frequency can be reduced (e.g., event-triggered sampling)
(2) The communication requirements are less at the steady state, (e.g., lower priority can be assigned to the feedback signals)
Sensitivity of control performance depends on the state of the controlled plant
Control Tasks - Characteristics
Technische Universität München
Bottomline
Real-time Systems Meeting deadlines is the center of attraction
CPS View Deadline takes the back seat As a result, the design space becomes bigger Resulting design is better, robust, cost-effective …
20
Technische Universität München
What about NCS?
Take network characteristics into account when designing the control laws Packet drops, delays, jitter …
21
Plant
T3 T4
Sensor
T1
T2
m1
m2 m3 +
- Controller
Network
Networked Control Systems
Technische Universität München
What about NCS? Answer: ANCS
ANCS – We can design the network By taking into account control performance constraints
Problem: How to design the network? Given a network, how to design the controller?
NCS problem
CPS Problem: How to design the network and the controller together? Co-design Problem
22
Plant
T3 T4
Sensor
T1
T2
m1
m2 m3 +
- Controller
Network
Arbitrated Networked Control Systems
Technische Universität München
CPS-Oriented Design Example
Control over FlexRay FlexRay is made up of time-triggered (TT) and event-
triggered (ET) segments Conventional design
Use time-triggered segment for control messages – one slot for each control message
Expensive
23
Challenge: Can we achieve the same control performance
with fewer time-triggered slots?
Technische Universität München 24
FlexRay – Event Vs Time-triggered
task
message
task x
Asynchronous
task
message
task
Synchronous
Jitter, lost/unused data
Predictable
x
Technische Universität München
Quality of Control vs. System State
The performance of a control application is more sensitive to the applied control input in transient state compared to that in steady-state
ET communication for the control signals is good enough in the steady-state
TT communication is better suited for transient state
25
r
t
Transient State Steady State
Observations
Settling time
Technische Universität München
Mode Switching Scheme
26
Plant in Steady-
State
Plant in Transient
Mode Change Protocol
Plant in Transient
Plant in Steady-
State
TT : ET : Schedule : Schedule :
Controller : Controller :
SissSi
t r
K it r
K iss
M ssM t r
x[k]T x[k] > Et r
Mode Change Request After
After dwell time Tidw
Technische Universität München
Example
We consider two distributed control applications communicating via a hybrid communication bus
27
We apply state-feedback controller for both, i.e., u[k] =Fx[k]
Technische Universität München
Performance with TT Communication
28
Control Gains
Quality of control
Converges very fast without any oscillation
C1
C2
Technische Universität München
Performance with ET Communication
29
Control Gains
Quality of control
C1
C2 Large oscillations and long settling time
Technische Universität München
Performance with Switching
30
Quality of control
C1
C2 Performance is better than that with ET communication but we consume less TT communication slots
We have one shared TT communication slot. The control messages are transmitted via ET communication when they are in steady state and switches to TT communication when transient state occurs due to some disturbance
Technische Universität München
Experimental Setup
31
Technische Universität München 32
Mixed time-/event-triggered Purely event-triggered Purely time-triggered
Experimental Results
Technische Universität München
Concluding Remarks
33
Viewing cyber-physical systems not as traditional real-time systems Breaks traditional concepts of abstraction? Current practice:
Specify control laws Specify architecture and schedules Test and debug (modify architecture and schedule)
Problem: Synthesize the architecture/schedule from control performance constraints Large constrained optimization problem
Technische Universität München 34
… and now something more futuristic