Abstract: To solve the problems associated with the traditional rapidly-exploring random trees (RRT) algorithm, including the lack of orientation and the slow convergence speed in the motion planning of the robot arm, we propose an extension-point selection strategy and adaptive step-size strategy based on the traditional RRT. In addition, when the algorithm is being trapped by a local minimum, it adopts an avoidance regression mechanism to quickly remove the minimum. Lastly, the Dijkstra algorithm is used to optimize the path generated by the improved algorithm, and the effective path at the end of the manipulator is transformed into an optimal pose path by the manipulator planning module. The improved algorithm and other algorithms are simulated in Matlab and ROS. The experimental results show that the proposed algorithm can effectively guide the direction of growth of the RRT tree, prevent falling into a minimum value, and improve the convergence speed of the algorithm and the motion planning efficiency of robot arm in simulation.