GEM UNIT Autonomous GEM Vehicle

1
GEM UNIT! Autonomous GEM Vehicle

• Adrian Afan
• Thomas Fong
• Martin Lizarde Jr.
• Zaynal Pham
• June 2nd, 2005

2
Introduction

• Objective
• to modify a neighborhood electric vehicle (GEM)
  to autonomously move from one location to another
  while detecting and avoiding objects along the
  way
• the GEM can be manually controlled via a joystick
  in a drive-by-wire configuration
• when the desired location is reached, the GEM
  will park itself in a designated parking spot

3
Introduction (Continued)

• Approach
• the GEM is guided using x and y mapping via
  Global Positioning System (GPS) and uses a laser
  range finder to detect and avoid obstacles
• parking will be implemented by utilizing a
  specifically designed object that the vehicle
  will detect and use as a reference point to align
  and park directly in front of
• The joystick is interfaced to a microcontroller
  to control the vehicle manually

4
Introduction (Continued)

• The microcontroller will be used to control the
  throttle, braking, steering based on the input
  from the joystick
• The master controller (laptop) will be used for
  more demanding applications and processing such
  as communication with the GPS, laser range
  finder, digital compass
• It processes the data to find the desired heading
  of the vehicle toward the destination point and
  the path needed to avoid any obstacles on the way

5
Technical Approach
6
Circuit Schematic
7
Microcontroller Algorithm

• Acceleration
• Reacts Linearly to the Joystick Input
• Joystick increases acceleration when pushed
  forward
• Speed range is 0 to 15mph
• Braking
• Joystick increases braking when pulled backward
• Can stop vehicle slowly or almost instantaneously

8
Microcontroller Algorithm - Turning

• Turning
• Joystick is used as input, steering encoder used
  as feedback
• Both cover range from 0 to 5V, left to right
• Voltages are compared for turning
• If encoder has smaller voltage, wheels are
  signaled to turn right
• If encoder has larger voltage, wheels are
  signaled to turn left

9
Microcontroller Algorithm - Communication

• Serial Communication
• If data is detected on the serial port and the
  joystick is in the neutral position, the vehicle
  will run autonomously
• Serial data is ignored if the joystick is outside
  of the neutral position
• ASCII value determine the actions of the GEM
• Serial data need to be sent in pairs
• Serial timeout present to prevent microcontroller
  from freezing while waiting for message

10
Autonomous Algorithm

• System designed to home in on final point
• Vehicle will go a constant speed unless no paths
  are detected or it has reached the final
  destination
• Algorithm
• If ( vehicle is close to the desired location)
• Tell the vehicle that it has reached the final
  destination and stop the vehicle

11
Autonomous Algorithm - Continued

• If ( Destination is almost directly North or
  South of vehicle)
• Assign a predefined desired angle
• Else
• Convert the GPS coordinate system into a regular
  XY coordinate system
• Then the desired angle is calculated using

12
Autonomous Algorithm - Continued

• Desired angle is converted into the regular
  coordinate system
• Heading angle is converted into the regular
  coordinate system
• The difference in the desired angle and the
  heading angle is determined
• From this difference, the bias path is returned
  to the obstacle avoidance algorithm
• There are 5 predetermined paths the vehicle can
  take.
• The bias path is the path toward the destination
  without accounting for obstacles

13
Obstacle Avoidance Algorithm
14
Testing Plan

• Demonstrate manual GEM control via joystick
• Tweak microcontroller code for functionality and
  ride smoothness
• Ensure complete safety of system through testing
  of responsiveness as well as emergency cut-off
  and stopping procedures.

15
Testing Plan - Continued

• Demonstrate the GEM moving from one location to
  another at the CE-CERT parking lot while
  detecting and avoiding objects
• Create obstacle map while under manual control
• Find current GPS location and path to desired
  location while under manual control
• Allow vehicle to navigate autonomously to desired
  location via GPS without obstacles.

16
Testing Plan - Continued

• Integrate obstacle map with path planning
  algorithm and find paths around obstacles to
  desired location while under manual control.
• Full autonomous movement to desired location with
  obstacle detection and avoidance
• Input multiple destination points into program.
  Vehicle should drive to each point, stop
  momentarily and continue to the next point

17
Results

• Mechanical Implementation Successful
• Braking, Throttle and Steering respond correctly
  to joystick inputs
• Steering Motor Mount does not flex while in use.
  Chain and sprocket are very durable and have no
  slack to add to the steering delay
• Emergency Brakes have been tweaked to stop the
  vehicle as fast as possible

18
Results - Continued

• PC Communication with sensors and microcontroller
  work well
• Laser range finder updates the obstacle map at 1
  Hz
• GPS and Compass values are updated whenever the
  laser range finder updates
• Vehicle runs autonomously with little difficulty
• Vehicle detects obstacles and can maneuver around
  them to follow a clear path
• Vehicle stops upon nearing the destination

19
Results - Continued

• Minor Issues
• Brake controller has a threshold voltage before
  it activates the brakes.
• The joystick needs to be pulled back about a
  third of the way
• Causes the braking to be very sudden
• The steering is slow
• Takes about 2-3 seconds to get from one side to
  the other
• Laser Range Finder can only detect objects higher
  than 3 feet tall
• Can be blinded by bright sunlight at the right
  angle

20
Future Work

• Improvements to Design
• Implement Reverse and other car controls into
  Microcontroller
• Add ability for microcontroller to control
  direction
• Allows vehicle to back out of traps
• Buttons on Joystick can be mapped to horn,
  changing vehicle direction, or signaling
• Add more safety features
• Throttle cutoff if power to DAC circuit is lost
• Component Cutoff switches near driver
• Toggle brake lights when car is stopping

21
Future Work - Continued

• Parking
• Identify special obstacle and park in front of it
• Update Faster
• Stabilize the code for updating the obstacle map
  at 5Hz
• Creation of PCB for DAC circuit
• PCB has already been designed
• Need to fix pin assignments
• Improved Path Planning and Obstacle
  Identification Algorithms

22
Future Work - Continued

• Application for Future
• Identify an obstacle and follow it
• Expansion to larger vehicle and system to compete
  in DARPA Grand Challenge
• Next Slide Video
• Focus on the two stationary GEM vehicles
• Notice how it avoids the first one by turning
  right and then avoids the second one by turning
  left
• After the vehicle passes the obstacles, it
  realigns with the destination point and heads
  towards it
• Vehicle stops very close to the desired
  destination point marked by the cone

23
Video
24
Pictures

• (From Left to Right) Martin Lizarde, Thomas Fong,
  Adrian Afan, Zaynal Pham
• Special Thanks to Dr. Barth, Dr. Beni and all the
  people who helped us with this project!
  G-g-G-gem unit

25
Appendix A - Budget
26
Appendix B - Cost Analysis/Alternate Solutions

• Electro-Hydraulic Brake Controller
• While it stops the vehicle very well, it is not
  smooth as originally assumed to be
• Possibly because a dedicated brake controller
  with an internal accelerometer was not purchased
  with the system
• Could be replaced by a linear motor operating the
  master cylinder
• This would be smoother, but would need feedback
  to get proper braking

27
Appendix B - Cost Analysis/Alternate Solutions
Continued

• SICK Laser Scanner
• Provides a very accurate map of surroundings that
  it can see
• Does not see small objects that fall underneath
  its field of vision
• Needs to be supplemented by ultrasonic sensors
• Cannot be replaced by ultrasonic sensors because
  they lack the range given by the laser scanner
• Can be replaced by computer vision
• Faster computer would be needed with good
  obstacle detection algorithms

28
Appendix B - Cost Analysis / Alternate Solutions
- Continued

• Steering Motor
• Faster Motor with Similar Torque
• This would increase responsiveness of steering
• Small amount of space available around mounting
  point
• Could be replaced with Linear Motor
• Would need to be very strong to move tires
• Novel mechanical mechanism could multiply force
  exerted on tires and ensure that the physical
  limits of steering mechanism are not violated