An improved RRT-Connect algorithm, which synthesizes the merit of Dijkstra policy, is proposed to boost its prior one in view of the disadvantages of node extension inef?ciency, poor safety and multiple invalid path points in the obstacleavoidance path planning of spatial manipulators. Firstly, the path search efficiency and safety of the traditional RRT-Connect algorithm are improved by the presenting of dual trees to extend the target change strategy, extreme greed strategy and new collision detection method. Secondly, the optimization feature of the Dijkstra algorithm is utilized to eliminate redundant invalid nodes in the planned path to achieve the effects of high search speed, short path length and high security in path planning. Finally, algorithmic simulation experiments were performed through the MATLAB Robotics Toolbox, which showed the effectiveness and feasibility of this method.