CajunBot-II: An Autonomous Vehicle for the DARPA Urban Challenge

CajunBot-II is a 2004 Jeep Wrangler Rubicon equipped with an INS/GPS for localization. It has is equipped with a variety of environment sensors, including SICK LMS LIDAR sensors, Alasca XT LIDAR sensors from Ibeo, stereo vision, and radars. The software system is decomposed into a collection of programs communicating via a blackboard supported by a layer of middleware. Localization from the GPS and INS is improved by a Kalman filter based algorithm using a SICK LIDAR, and vehicle commands. Environment sensor data is fused using probabilities derived from the operating context. The mission is mapped into a series of maneuvers using the mission, vehicle state, and environmental information. A maneuver is applicable only on a path with a certain geometry. Execution of a maneuver takes into account the dynamic driving conditions. Human driving experiences are used to catalogue the maneuvers needed to satisfy the UC requirements. The team uses a 3-D physics based simulator for virtual testing. By using an RNDF file to create a world, the simulator provides a seamless way to perform the same tests in the field and in the lab. This compresses the debugging effort because scenarios observed in the field can be recreated in the lab.