Probabilistic Tracking and Planning in Ground Robots (ARO)
The goal of this research is to develop a unified theory for probabilistic perception and planning in autonomous ground vehicles, so as to enable reasoned, intelligent planning rather than simple reactive planning.
1. Sensor Fusion for Detection, Tracking and Classification: Develop a robust tracking system using additional sensory information (e.g. cameras), context from surroundings, and new methods such as negative information. More advanced perception, including classification and behavior estimation, will also be considered.
2. Probabilistic Trajectory Planning and Control: Develop trajectory planning methods that make direct use of probabilistic information evolving in real time from Task 1. Approaches include on-line optimization methods that include contingency planning.
3. Theoretical, Simulated, and Empirical Analysis and Verification: Analyze and verify the theoretical approaches using simulation and Cornell's autonomous car Skynet.
Contingency planning over probabilistic hybrid obstacle predictions for autonomous road vehicles. IEEE Transactions on Robotics [Internet]. Ieee; 2013;to be publ. http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=5652763 .