Optimization of Safety and Comfort in Car-following Scene Based on Reinforcement Learning

In order to ensure the safety and comfort performance in the car-following scene, based on the existing research, this paper chooses to use a deep reinforcement learning algorithm to build the following model and explore its performance in safety and comfort. Our research contents are as follows : (1) summarize the existing research about car-following, and summarize the principles of the existing vehicle following models; (2) do research on car-following safety and comfort index parameters; (3) use the principle of Deep Deterministic Policy Gradient (DDPG), and build up our car-following model based on DDPG algorithm; (4) build the car-following simulation environment in SUMO platform and train the model; (5) for testing our model, we use Intelligent Driver Model (IDM) for comparison and analyzed the simulation result from safety and comfort. Finally, it is concluded that our model can further improve vehicle comfort while ensuring safety. The research of this subject is helpful to the safe and comfortable driving in the car-following driving scene, and has certain significance and practical value in further improving the automatic driving technology of the vehicle.