Aiming at the problems of slow motion planning, low efficiency, and high path cost of multi-degree of freedom picking manipulators. The Cauchy target gravitational bidirectional RRT*(CTB-RRT*) algorithm was proposed. Heuristic sampling was carried out by Cauchy distribution method to reduce the blindness of sampling. Through introducing target gravity, the step length of the random growth direction and the target direction were dynamically adjusted to improve the local search speed. What’s more, a node rejection strategy was introduced to eliminate unnecessary sampling nodes and improve calculation efficiency. Through two-dimensional and three-dimensional algorithm simulation experiments, the 6-DOF manipulator's obstacle avoidance picking simulation experiment verified the effectiveness of the proposed algorithm. The results of research and simulation experiments showed that the path cost of the improved algorithm was reduced by 5.5%, the search time was reduced by 71.8%, and the number of sampling nodes was reduced by 64.2%, comparing with the RRT*-connect algorithm. To verify the feasibility of the algorithm, the 6-DOF manipulator was controlled to perform obstacle avoidance picking movement in the robot operating system (ROS). The success rate of motion was 99%, and the running time was 0.33s. Comparing with the RRT*-connect algorithm, the path cost was reduced by 12.6%, and the running time was reduced by 69.2%. The number of expansion nodes was reduced by 76.3%.