Aiming at the problems of low efficiency and poor accuracy of RRT*-Connect algorithm in manipulator motion planning, a heuristic RRT*-Connect manipulator motion planning algorithm with adaptive step size was proposed. Firstly, the target bias strategy was introduced to sample with the ellipsoid subset constraints, so that the sampling points can converge to the optimal value more quickly. Secondly, as the nodes were extended, an adaptive step size strategy was designed to reduce the number of iterations of the algorithm, effectively reduce the running time and shorten the length of the planned path. Then, when the total number of nodes in the search tree was greater than the preset threshold, the search tree was pruned by the search tree optimization pruning method, and the invalid sampling points were deleted to further reduce the running time. In order to verify the advantages of this algorithm, Matlab simulation comparison was carried out with RRT*, RRT* -Connect and IRRT* algorithms in various planning scenarios. The simulation results showed that the proposed algorithm had faster convergence speed, higher accuracy and efficiency in the planning process. In order to verify the practicability of this algorithm, different obstacle experimental scenes were constructed, and the experimental verification was carried out on Sawyer manipulator experimental platform. The experimental results showed that this algorithm had strong adaptability in different obstacle environments.