This thesis presents an obstacle avoidance strategy designed to meet the demands of the 2007 Defense Advanced Research Projects Agency (DARPA) Urban Challenge. The central theme of this work is the combination of the reactive and the global motion planning techniques to achieve an efficient and robust obstacle avoidance module (OAM). The aim of the OAM is to navigate the autonomous ground vehicle (AGV) robustly and without any collisions in an outdoor urban environment using the local sensory information. The OAM attempts to alleviate the problems faced by purely reactive approaches and provides an obstacle-free navigation in unknown and complicated environments. The motion commands generated by the OAM are: a smooth collision-free path constructed using the Cubic Bézier Polynomials and a longitudinal speed at which the path needs be followed. Furthermore, the OAM is enhanced by the capabilities which allow the AGV to navigate safely among the moving traffic.
The proposed OAM was tested in simulation extensively to validate its capabilities and to improve its overall performance. A number of challenging outdoor obstacle fields were constructed to test the functionalities of the OAM. The OAM was successfully integrated on the Ohio State University Autonomous City Transport (OSU-ACT) vehicle which competed in the 2007 DARPA Urban Challenge. Finally, the OAM demonstrated its reliability and the robustness by completing all the given tasks successfully in the National Qualification Event (NQE) of the Urban Challenge. The main focus of this thesis is on the overall algorithm structure and the demonstration of its capability using simulation and the experimental results. Furthermore, the challenges associated with the motion planning of nonholonomic robots, with sensor limitations and the external disturbances are discussed.