SIPHER students: Jessica Kane and Thao Nguyen

1
Model-Based Autonomous Car Controller Design
Results The model-based controller enabled the
car to follow various trajectories autonomously.
In the following figures, the green line is the
desired trajectory the blue line is the cars
actual position.
Abstract A small radio-controlled car was
equipped with sensors enabling it to localize
itself. Using the Vanderbilt Embedded Computing
Platform for Autonomous Vehicles (VECPAV)an
existing infrastructure designed for autonomous
helicopter flighta PD-controller was developed
to induce the car to autonomously travel a given
set of trajectories.
Approach

• The VECPAV computing platform has been
  constructed to allow for system design in
  Simulink during design time and for automatic
  C-code generation and distribution onto real-time
  QNX computational nodes.
• Closed-Feedback Hardware Loop
• Car sends its (x, y, z-rotation) position to
  tracker
• Tracker transmits information to the Boxx
  processor
• Data is processed through the controller
• Steering and throttle signals are transmitted to
  real-time nodes
• Radio transmitter receives the signals and sends
  them to the car

Figure B Straight Line Trajectory The stepping
characteristic is caused by the cars motor. Its
slowest speed is faster than the trajectory. The
car compensates by stopping before it gets ahead
of the trajectory. Motion resumes when the car
has fallen behind the trajectory.
Figure C Counter-clockwise Circle
Trajectory Here, the car completes eight complete
counter-clockwise circles. The car remains
consistently behind the trajectory because the
speed controller is designed to move the car only
when it is more than 200 mm from its desired
position.

• Real-time Embedded Software System
• The Simulink model run by RT-Lab is composed of a
  loop
• between the console and the controller.
• In the console
• Trajectory outline
• Scopes for viewing real-time data
• Constants to be adjusted in real-time
• In the controller
• Data received from the console
• Errors calculated

Figure D Racetrack Trajectory Designed to
determine how the controller handles combinations
of straight-line and circular motion.
rreference (trajectory) coordinates m
measured (tracker) coordinate
Motivation Autonomously driven vehicles are no
longer a dream for the future, but a reality in
the present. Researchers have successfully
developed self-driven, full-size automobiles.
Model-based experimentation with controllers for
an autonomous remote-controlled vehicle continues
this research on a smaller scale.
Figure E Figure-8 Trajectory Designed to test
the controllers response to a combination of
diagonal lines, clockwise semicircles, and
counterclockwise semicircles.
The throttle controller (6) is a simple
P-controller. The steering controller (8) is a
PD-controller for higher accuracy.

• Future
• Incorporate second car and/or helicopters
• Develop more accurate model of the car for error
  calculations
• Design more robust, adaptive speed controller

Figure A Tracker Coordinate System The rooms
coordinate system includes a transition point
from 180 to -180. It caused discontinuity in
the heading error and inconsistent behavior in
the cars motion. The final solution renormalizes
heading error (4) from -90 to 90.

• SIPHER students Jessica Kane and Thao Nguyen
• Graduate Student Advisors Graham Hemingway,
  Peter Humke