Trajectory planning for robotic manipulator based on improved DDPG algorithm

Deep reinforcement learning (DRL) algorithms are increasingly applied to robotic manipulator planning. However, conventional DRL approaches suffer from slow learning convergence and low success rates in industrial trajectory planning tasks. To address these issues, this paper proposes an improved deep deterministic policy gradient (DDPG) algorithm that more effectively achieves time-optimal trajectory planning for robotic manipulators. Firstly, a radial basis function neural network is introduced to calculate nonlinear function values during parameter training to improve the learning convergence speed of the algorithm. The gradient descent algorithm is used to update the weights of the neural network. Meanwhile, the SumTree sample pool is used to screen high-quality samples and improve the utilization rate of the algorithm. The simulation experimental results show that compared with the traditional DDPG algorithm, the improved DDPG algorithm proposed in this paper has the torque and angle of each joint of the robotic manipulator change steadily, which improves the utilization rate of the algorithm in trajectory planning and the learning efficiency of the strategy.