Poster 25: Cooperative Short-horizon Path Planning for Autonomous Vehicles utilizing Actor-critic Reinforcement Learning Approach
(Room Palms Ballroom)
16 Jul 19
5:30 PM
-
7:00 PM
Tracks:
Vehicle Automation Technology
We tackle the problem of coordination among autonomous and human-driven vehicles in challenging driving maneuvers such as crossing an intersection, merging to a platoon, and exiting e freeway. We leverage the vehicle-to-vehicle (V2V) communication to enable vehicles to share their intent and planned path. Our proposed method relies on an actor-critic reinforcement learning based approach to optimize the shared information among vehicles enabling them to react and adapt to each other in a cooperative manner. We evaluate the proposed algorithms in our cooperative vehicle simulator, Co-Sim, with hybrid driving scenarios in which human drivers and self-driving cars co-exist.