A Multi-Task Fusion Strategy-Based Decision-Making and Planning Method for Autonomous Driving Vehicles

The autonomous driving technology based on deep reinforcement learning (DRL) has been confirmed as one of the most cutting-edge research fields worldwide. agent is enabled to achieve goal making independent decisions by interacting with environment and strategies feedback from environment. This widely used in end-to-end tasks. However, this field faces several challenges. First, developing real vehicles expensive, time-consuming, risky. To further expedite testing, verification, iteration algorithms, a joint simulation development validation platform was designed implemented study VTD–CarSim Tensorflow framework, work conducted platform. Second, sparse reward signals can cause problems (e.g., low-sample rate). It imperative for be capable navigating an unfamiliar safely under wide variety weather or lighting conditions. address problem poor generalization ability unknown scenarios, deterministic policy gradient (DDPG) decision-making planning method proposed accordance multi-task fusion strategy. main task DRL auxiliary image semantic segmentation were cross-fused, part network shared reduce possibility model overfitting improve ability. As indicated experimental results, first, built exhibited prominent versatility. Users easily substitute any default module customized algorithms verify effectiveness new functions enhancing overall performance using other modules strategy competitive. Its better than certain tasks, which improved vehicle algorithm.