Abstract:〖WTBZ〗In order to solve the problem of multi-rotor UAV quickly avoiding obstacles during the inspection process of industrial gas pipelines, an improved algorithm based on Rapidly-exploring Random Tree (RRT) is proposed, aiming at the problems of high randomness, slow convergence speed and long and tortuous planning paths of traditional RRT algorithms, combined with the characteristics of the space environment. Two aspects including random point sampling and path optimization have been improved. First, this paper sets the value range of random sampling points, then introduces the target bias sampling strategy to guide the expansion direction of the random tree, and finally cuts and smoothes the generated path. The simulation experiment is compared with traditional RRT algorithm and heuristic RRT algorithm with path correction in different digital analog maps. The experimental result shows that the improved RRT algorithm can quickly generate a short and smooth obstacle avoidance path when performing gas pipeline inspection tasks.