Path Planning in Urban Environment Based on Traffic Condition Perception and Traffic Light Status

Vehicle path planning problems have been studied for decades. The existing path planning methods are suitable for simple objectives. However, for complex tasks such as planning paths for vehicles considering the effects of pedestrians, traffic lights, etc., it is difficult to design a reasonable cost function for the deterministic algorithm or a reasonable heuristic function for the heuristic algorithm. In this paper, we proposes a path planning model based on traffic light status and traffic condition awareness. When a vehicle arrives at a new road section, it senses the traffic light status, distribution and vehicle positions in the road network on each road section through V2V and V2I communication, and based on this information, we use an A2C-based deep reinforcement learning method to dynamically plan the shortest path for the vehicle in real time. Experiments show that the proposed method works effectively in terms of saving on driving time and waiting time to reach any destinations, compared to the existing solutions.