Implementation of kinematic control for a DARPA Urban Challenge vehicle

The DARPA Urban Challenge provided a purpose for autonomous ground vehicle research in recent years. The University of Utah entered the competition with a vehicle named Red Rover. In the course of preparing for the competition a path following control technique was developed and implemented, modified from a controller found in the literature. The implementation is a kinematic controller with a feed forward term, gain scheduling and safety check, termed the error governor. Results show the necessity of the gain scheduling and governor to maintain stability through a range of operational conditions. Postcompetition efforts uncovered a saturated communications channel, whose repair greatly improved tracking performance. In addition actuator feedback was restored, which afforded the implementation of an extended Kalman filter. This filter successfully accomplishes sensor switching such that sensor measurements can be ignored when they contain false information. The filter implementation successfully smoothes state estimates and thereby achieves the underlying goal of giving the system a natural steering motion for the benefit of the passengers. An initial effort to make the onboard wheel encoders a reliable source of velocity data is also explored. However, given an eight-spoke hardware design, an acceptable solution was not found.