EE192 Spring 2009 – Final Presentation

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 1

Farzad Fatollahi-FardHartej DhamiAman Mouhidin

Overview Hardware

General

Sensors

Software

Controls

Performance

Roles and Contributions

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 2

Hardware - General We used ADUC7026 evaluation board

ProsWe had cross-talk in adjacent ADC pins, with 11 ADCs we were able to

use every otherEasy to change GPIO functions (All pins have headers)

ConsAccommodate larger board (5-6 times more area than mini-board)Always get asked “Why is your board so big?” We were assigned the ADUC7026, we just stuck with it.

We used hi-torque instead of hi-speed servo Pros

Turns not limited by lateral momentum

ConsLower slew rate (0.08 rad/s vs. 0.06 rad/s)

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 3

Sensors We used three sensors

Middle, Left, and right

PCB can handle four sensor inputs if need be

We used a two stage amplifier One gain stage and One buffer stage

Our power circuitry and sensor circuitry had common ground plane We did not run into any noise problems with a common

ground plane

The DC boost and sensor op-amps were on opposite parts of the PCB

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 4

SensorsCharts

-10

-5

0

5

10

-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8

Estimate

Estimate

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 5

0

500

1000

1500

2000

2500

3000

3500

4000

4500

Left

Right

Center

Software Optimizations

We tried to do bit shifting if possible

Common conditions are at top of if statement tree

Used shorts instead of integers

There is no ‘special’ feature

We believed that simpler code would result in best performance (KISS, more knobs…)

Able to change kp & kd via Bluetooth™

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 6

Software Locked in last turn direction to recover if track is lost

Used saturation on servo turn angle calculation to protect against large current draw from V-REG

Used timers for servo PWM and in-built PWM for motor

Used version control for code (SVN)

Better is not the enemy of good with version control for code anymore

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 7

Control We used a PD controller

centerServo + (kp × diff)/100 + (kd × (diffL - diff))/100;

We started with an estimate for the kp and kd values Then we changed kp and kd via Bluetooth™ until we saw a stable result Initial Problems:

Initially we were sampling the ADC values in the while loop, every xth

iteration of the loop Our car was very wobbly and this was because we were not sampling

often enough We changed our code to sample ADC values based on Timer2 interrupt

Simulations proved helpful in understanding the concepts. We tried doing simulation to get a starting value for kp and kd, those

values were close but a little off from what we ended up using kp: 5.54 rad/m; kd: 1.82 rad × s/m

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 8

Performance Runs well except following short comings:

Twitches on straight-ways at crossings

Slips on turns

Tries to run away at 100% Duty Cycle PWM when touched inappropriately

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 9

-10

-5

0

5

10

15

20

0.0

00

.06

0.1

10

.17

0.2

30

.28

0.3

40

.40

0.4

60

.51

0.5

70

.63

0.6

80

.74

0.8

00

.85

0.9

10

.97

1.0

21.

08

1.14

1.20

1.25 1.31

1.37

1.4

21.

48

1.54

1.59

1.6

51.

711.

761.

82

1.8

81.

94

1.9

92.

05

2.11

2.16

2.22

2.28

2.33

2.39

Err

or(

cm)

Error(cm) VS Distance(m) @2m/s

Error

Things we tried Use cross sensor

If cross value high enough ignore other sensors and keep last servo value

Decreased performance

Car twitches a lot more.

Speed encoder

Simple addition/subtraction scheme to modulate speed

Held speed within 0.5 ft/s of requested speed

Worked fine indoors

Did not work in sunlight

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 10

Roles and Contributions Hartej Software Development

Car platform/boom design

PCB Schematic & Layout 2

Farzad Software Development

PCB Schematic & Layout 1 and 2

Aman Circuit Design

PCB Schematic & Layout 1 and 2

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 11

Fin

1/22/2010University of California, Berkeley - Electrical Engineering and Computer Science 12