Planning and Control for the Darpa Urban Challenge by Yoshi Kuwata, Chun Sang Teo and Gaston Fiore

DGC 2007 MIT
The 2007 Darpa Grand Challenge is the third in a series of competitions organized by DARPA to accelerate research and development of unmanned robotic vehicles capable of performing military supply missions. The 2007 DGC is the first such competition that requires the vehicles to navigate an urban environment. Some challenges arising from such a setting include the requirements to: i) avoid both static and moving vehicles/obstacles, ii) plan feasible trajectories in real-time, iii) obey rules of the road, iv) plan trajectories that perform 3 point turns, parking in potentially highly constrained parking spots, etc. Team MIT participated in the 2007 DGC and was placed fourth among six teams that completed the race on November 3rd, 2007. ACL faculty and students were involved in the design and development of the Rapidly-exploring Random Tree (RRT) based motion planner and vehicle controller. Our approach is unique in that a single unified planner was used, whether operating in a highly constrained structured lane following scenario or in the unstructured parking lot scenario (which necessitates a highly exploratory planning approach).

Related publications