UNMANNED DRIVING PATH TRACKING CONTROL BASED ON THE IMPROVED DEEP DETERMINISTIC POLICY GRADIENT

In view of the shortcomings of path tracking accuracy and poor driving stability of driverless cars at high speed, a driverless path tracking control strategy with improved deep deterministic policy gradient (DDPG) algorithm is proposed. By designing a state quantity optimizer to eliminate the less influential state quantities and optimizing the gradient policy learning efficiency, the process of policy learning weights was thereby improved to enhance the accuracy of the tracking path and achieve stable and accurate tracking of the track route by driverless cars. The experimental results show that compared with the Deep Q Network (DQN) algorithm and the DDPG algorithm, the proposed algorithm can enable the driverless car to track the route quickly and improve the tracking control accuracy effectively, which verifies the tracking performance of the proposed algorithm.