UAV 3D online track planning based on improved SAC algorithm

Online track planning technology is a key technology to improve the survivability and intelligence of UAVs. In this paper, a Soft Actor-Critic (SAC)-based online track planning algorithm is proposed by combining Artificial Potential Field (APF) algorithm and self-attentive mechanism. First, for the characteristics of high-dimensional complexity of environmental states presented in the 3D environmental space, the self-attention mechanism is introduced in the Actor network of SAC algorithm in order to analyze and process the state information. Secondly, the artificial potential field method is combined with the reward function of SAC algorithm to solve the problem of reward sparsity in online track planning. Finally, the observation, action, and reward functions of the algorithm are designed according to the characteristics of UAV online track planning. The experimental results show that the algorithm convergence speed and success rate of the improved SAC algorithm are significantly improved, which can realize the 3D online track planning of UAVs.