It is challenging to develop an online path planning algorithm for self-driving (Ackermann-steering) vehicles to find collision-free and kinematically-feasible paths, that is efficient and safe for various dense environments. A possible solution is the combination of a motion planning algorithm and a deep reinforcement learning (DRL) trained policy. We propose a kinematically constrained RRT-based path planning algorithm integrating with a trajectory parameter space (TP-space) with three novel improvements to meet the above requirements. We also propose our experiences in training a robust path planning policy by DRL. We evaluate the methods in various environments. The experimental results show that the methods achieve competitive performance compared with the state-of-the-art.
Professor, School of Computer Science and Technology
University of Science and Technology of China
Jianmin Ji is an Associate Professor at University of Science and Technology of China (USTC). He received a B.Sc. degree in Computer Science and a Ph.D. in Computer Science from USTC in 2005 and 2010. From 2010 to 2012, he was a Postdoctoral Fellow with the Department of Computer Science and Engineering, The Hong Kong University of Science and Technology (HKUST). In 2012, he returned to USTC as an Associate Professor. He was a Visiting Scholar with Griffith University in 2015, University of Alberta in 2016, and Carnegie Mellon University in 2017. He was the Senior Program Committee Member of IJCAI'15, '17, '20, '21, Co-Chair of the 2015 RoboCup Symposium, and Co-Chair of the Workshop on Nonmonotonic Reasoning, Action and Change (NRAC-13) in IJCAI-13. He is interested in AI and robotics, especially in self-driving, reinforcement learning, robot navigation, knowledge representation and reasoning, and answer set programming. He designed the cognition component of KeJia and JiaJia robots, which was considered as one of the "Best Technique Solutions for Cognition" in Artificial Intelligence 2015.
