Aiming at the problems of rapidly-exploring random tree (RRT) algorithm, such as blind generation of search nodes, non-directional expansion and tortuous path, the traditional RRT algorithm is improved. The probability value is introduced to reduce the number of random points. The target bias strategy is used to guide the random tree to grow towards the target point. The variable step size is used to reduce the oscillation near the target point. The path is simplified and the initial path is optimized by cubic B-spline curve. The simulation results show that the improved RRT algorithm effectively shortens the path length, reduces the number of turning points, improves the smoothness of the path, and has good effectiveness.