An end-to-end autonomous driving system in a strong interaction scenario considering the style of other vehicles

Abstract

This invention pertains to autonomous driving systems for automobiles, specifically an end-to-end autonomous driving system that considers the driving styles of other vehicles in highly interactive scenarios. An image feature extraction module extracts image features; a traffic participant style perception module extracts the driving style features of traffic participants; these features are concatenated and input to a value estimation module and a policy module. The value output by the value estimation module guides the training of the autonomous driving policy in the policy module, and the actions output by the policy module are used for the interaction between the autonomous vehicle and the traffic environment. This invention utilizes implicit style features as a branch of the model input, which can express prior cognition, thereby accelerating the training efficiency of the reinforcement learning agent and improving the agent's ability to learn and process vehicle-to-vehicle interaction relationships. Simultaneously, the reinforcement learning reward is dynamically adjusted according to the driving styles of traffic participants, enabling the agent to adopt different response strategies based on the different styles of traffic participants in highly interactive scenarios.

Images