Human-like longitudinal velocity control based on continuous reinforcement learning

Traditional intelligent driving systems suffer from low efficiency in the dynamic traffic environment because of their rigid planning modules. On the other hand, experienced human drivers can deal with dynamic situations adaptively without a complex planning system. This study aims to develop a learning-based driving system that can learn from human drivers and realize the human-like longitudinal control. The proposed system contains two main parts: a reinforcement learning module for learning human driving strategies and a PID control module for converting the strategies learned to control actions for vehicles. Experiments based on simulation are carried out to test the performance of the proposed system. A driving simulator based on the software PreScan is used to collect the driving data from human drivers and build the test scenarios. Experimental results show that the learning-based system can duplicate human driving strategies with acceptable errors in several predefined cases.