Aiming at the path planning needs of a seven-degree-of-freedom robotic manipulator in robot-assisted total knee arthroplasty, this paper proposes a heuristic sampling-based robotic arm path planning algorithm based on the traditional RRT* algorithm. Firstly, a feasible path is quickly expanded using a connect search strategy, the path is simplified by removing the redundant points of the initial path through the greedy strategy, the simplified path length is recorded to form the initial informed sampling region, and the gravitational gain coefficient is introduced probabilistically in the informed sampling, which further improves the convergence ability of the algorithm. Secondly, the optimization steps of re-selecting the parent node and rewiring the random tree in the dynamic range region are performed to gradually optimize the path length. Finally, the greedy strategy is used again to remove the path redundant points, and the quadratic Bessel curve is used to smooth the path. In order to verify the practicality of the proposed algorithm, different experimental environments are constructed using the Matlab platform, different algorithms are evaluated based on the TOPSIS entropy weighting method, and the comprehensive ability of different algorithms is compared in osteotomy environments through the ROS (robot operating system) simulation and experimental prototypes comparison experiments. The results show that the proposed algorithm is able to provide fast and effective path planning for the robotic manipulator in multi-scene.