MOBILE ROBOT PATH PLANNING BASED ON IMPROVED DQN ALGORITHM

It is important to guarantee real-time performance when the path planning is carried out in a dynamic and complex environment. Aimed at the overestimation problem and the slow convergence speed of DQN algorithm in mobile robot path planning, a C-RD3QN (combination-residual dueling double DQN) algorithm is proposed. Based on D3QN algorithm, the convolution layer was modified to the residual network structure, the action advantage function in the competitive network structure was used to estimate the action value function, and the state value function was combined with the reward value to achieve faster convergence speed. The simulation results show that the C-RD3QN algorithm can carry out better path planning.