Bianca WoodChris Culver

Shane ParkerYousef Al-Khalaf

Motivation

Challenge Our CapabilitiesSense of AccomplishmentSheer Fun

Objectives

Build a flying stable quadrotorAgileReal-time, intelligent decision-makingAutonomous

Challenges •Waiting for parts to come in •Not having a proper testing environment

•(worked in a ditch literally)•Making something fly, and also be stable• Having our frame break two days before our final presentation

Specifications

Each quadrotor is .91 m diamterHeight of .178 m Weight ~ 5 lbsAble to operate for 15 minutes on a single charge

A.I. Controller

Power System

Image Processor

Flight Control

Current Position

Shut Off

Memory Power Switch

Starting Pos.

Motor ControllerCamera

Block Diagram

Fire Control

A.I. Controller

Power System

Image Processor

Flight Control

Current Position

Shut Off

Memory Power Switch

Starting Pos.

Motor ControllerCamera

Block Diagram

Fire Control

A.I. Controller

Power System

Image Processor

Flight Control

Current Position

Shut Off

Memory Power Switch

Starting Pos.

Motor ControllerCamera

Block Diagram

Fire Control

A.I. Controller

Power System

Image Processor

Flight Control

Current Position

Shut Off

Memory Power Switch

Starting Pos.

Motor ControllerCamera

Block Diagram

Fire Control

A.I. Controller

Power System

Image Processor

Flight Control

Current Position

Shut Off

Memory Power Switch

Starting Pos.

Motor ControllerCamera

Block Diagram

Fire Control

Power

Micro Controller

Motor

Motor

Motor

Motor

Movement

A.I. Controller

Flight Control Sub-System

NavigationControl

Algorithm Coordinates /

Sensors Motors

Navigation•Coordinates come from AI computer

Stabilization•Readings come from sensors

PWM Signal

Rxacc = (ADCRx*Vref/1023 – Vzerog)/ Sensitivity

Ryacc = (ADCRy*Vref/1023 – Vzerog)/ Sensitivity

Rzacc = (ADCRz*Vref/1023 – Vzerog)/ Sensitivity

R2= Rxacc + Ryacc + Rzacc

2 2 2

-ADC = Value coming from accelerometer -Vref = Reference voltage from ADC -1023 = Max value of ADC bus-Vzerog = Acc under 0 g’s of force -Sensitivity = Relationship between changes in acceleration to change in output

Accelerometer

( ) Θxr = cos

-1 Rx

R

( ) Θzr = cos

-1 Rz

R

( ) Θyr = cos

-1 Ry

R

Gyroscope

θxy = Rotation around Z axis Yaw

θyz = Rotation around X axis Roll

θxz = Rotation around Y axis Pitch

Rate θxy = (ADCxy*Vref/1023 – VoltsZeroRate)/Sensitivity

Rate θxz = (ADCxz*Vref/1023 – VoltsZeroRate)/Sensitivity

Rate θyz = (ADCyz*Vref/1023 – VoltsZeroRate)/Sensitivity

-ADC = Value coming from gyro-Vref = Reference voltage from ADC -1023 = Max value of ADC bus-VoltsZeroRate = Output voltage when no rotation-Sensitivity = Change in output voltage with one degree per sec rotation

Combining Accelerometer and Gyroscope Data

• Takes Accelerometer data • Checks it against Gyroscope data and past output data• Corrects itself

• Rout(n) = Current output of Algorithm• Rout(n-1) = Last output of Algorithm • Rate θ = Gyro output • Rgyro = Current gyro & past output combined

Θxz(n-1) = atan2(Rxout(n-1), Rzout(n-1))Θyz(n-1) = atan2(Ryout(n-1), Rzout(n-1))Θxy(n-1) = atan2(Rxout(n-1), Ryout(n-1))

Θxz(n) = Θxz(n-1) + Rate θxz(n)*T

Θyz(n) = Θyz(n-1) + Rate θyz(n)*T

Θxy(n) = Θxy(n-1) + Rate θxy(n)*T

Rxgyro(n) = sin(Θxz(n)) / SQRT{1 + cos (Θxz(n) )*tan (Θyz(n))} 2 2

Rygyro(n) = sin(Θyz(n)) / SQRT{1 + cos (Θyz(n) )*tan (Θxz(n))} 22

Rzgyro(n) = SQRT( 1 – Rxgyro (n) – Rygyro (n))22

(T= sampling period)

• Rout(n) = Current output of Algorithm• Rout(n-1) = Last output of Algorithm • Rate θ = Gyro output • Rgyro = Current gyro & past output combined

Rout(n) = Racc + Rgyro ( ) w2

w1

( ) w2

w11 +

( ) w2

w1

*= How much to trust the gyro over the accelerometer

Computer VisionHaar Wavelets, first real time face detector.

Viola and Jones adapted idea, developed Haar-Like-Features. Considers adjacent rectangular regions at a specific location

in a detection window. Sums pixel intensities. Calculates difference between the sums.

Computer Vision•Integral Image Algorithm

•Single Pass Over the Image

•Evaluating any Rectangle in Constant Time

Overall Software Class Diagram

Artificial Intelligence Subsystem

State Machine

Navigation Mesh

Path With Navigation Mesh

Hardwarei7 Ivy Bridge16GB DDR3 1600 with 9-9-9-24 Timings120GB SSDNVIDIA 8900 GT

Voltage : 2.1 – 3.6

Frequency : 2.4 GHz

Data Rate : 250 Kbps

Range : 200 ft open space

Voltage : 5 v

Current : 500 mA

Power(2) Lithium-ion Polymer batteries11.1 v 2200mAhMounted on the

bottomCamera powered

by 9VPCB to disperse the

power

Power Distribution

Battery 1

Battery 2

PCB &

Voltage Regulator

ESC/Motor

ESC/Motor

ESC/Motor

ESC/Motor

PCB•Power in from left and right• Voltage regulator came with

MCU•5V Regulator•Receiving ~ 22V

• 2 diodes •4 arms to the ESCs/motors

PCB•Power in from left and right• Voltage regulator came with

MCU•5V Regulator•Receiving ~ 22V

• 2 diodes •4 arms to the ESCs/motors

PCB•Power in from left and right• Voltage regulator came with

MCU•5V Regulator•Receiving ~ 22V

• 2 diodes •4 arms to the ESCs/motors

PCB•Power in from left and right• Voltage regulator came with

MCU•5V Regulator•Receiving ~ 22V

• 2 diodes •4 arms to the ESCs/motors

FinancesMotors (2) $500 donationsPCB materials donatedSelf funded

BudgetItem Quantity

Price per Quantity Total Price

Quadcopter Frames 2 $30, $100 $130

Computer 1 $1,000 ($1,000)

Motors 12 $10 ($120)

Slowfly electric prop 1045R 4-piece set 4 $3 $12

Wireless Microprocessor 1 $150 $150

Wireless 2.4 GHz camera 1 $20 $20

Wireless 2.4 GHz USB receiver 1 $40 $40

Zigbee 1 $80 $80

ESCs 4 $7 $28

11.1V 2200mAh 25C LiPo Battery 2 $15 $30

PCB 1 - -

Miscellaneous Unknown UnknownAllocating

$150

Total $640

Thank youSponsors

Jeff MolerTim ParkerJohn Enander

Dr. Samuel Richie

Questions

Bianca WoodChris Culver

Shane ParkerYousef Al-Khalaf