In view of the problems of large search randomness, poor target bias and path tortuousness in the current robotic arm path planning algorithm, an APF-informed-RRT* algorithm based on bidirectional target bias was proposed. Firstly, probabilistic adaptive target bias strategy is introduced based on bidirectional informed-RRT* to reduce the randomness of search and improve sampling efficiency. Secondly, for path expansion, the artificial potential field method is integrated into the two-way search tree to reduce the number of iterations of the algorithm. At the same time, in the path growth stage, the dynamic step growth strategy is adopted, that is, the step size is dynamically adjusted according to the expansion trend of the search tree, so as to avoid local optimization and speed up the path search time. Finally, the redundant nodes are removed by the principle of triangle inequality, and then the path is smoothed by B-spline curve to obtain the optimal planning path. The simulation and comparison experiments with bidirectional RRT*, bidirectional informed-RRT* and bidirectional P-RRT* are carried out in 3D environment. Compared with bidirectional RRT*, the time is saved by 41% and the number of sampling points is reduced by 63%. Compared with two-way informed-RRT*, 58% less time and 68% fewer samples are collected. Compared with bidirectional P-RRT*, it saves 30% in time and 60% in sampling quantity.