To address the issues of low convergence efficiency, high path redundancy, and poor adaptability in traditional RRT algorithms for robotic arm path planning, an improved rapidly-exploring random tree(RRT) algorithm with multi-strategy fusion was proposed to enhance the path planning performance of robotic arms in complex environments. A six-degree-of-freedom robotic arm model was created. By combining goal-oriented and local obstacle-aware Gaussian mixture sampling mechanisms, the generation probability of sampling points was dynamically adjusted to enhance the algorithm's search ability in narrow channels and obstacle-dense areas. The k-dimensional tree(KD-Tree) structure was introduced to optimize the nearest neighbor search efficiency, an adaptive step size mechanism was designed to improve local passability, and a path pruning strategy was combined to reduce redundant nodes and enhance path smoothness. The trajectory was ultimately smoothed and generated by applying quintic polynomial interpolation in the joint space. The results verified by MATLAB simulation and robot operating system (ROS) experiments show that the average path length of the improved RRT algorithm is shortened by 13.2% to 18.3% compared with the traditional RRT algorithm, and the convergence time is reduced by 77.9% to 93.8%. In the physical experiment of the robotic arm, this algorithm also achieved smooth obstacle avoidance and continuous trajectory planning, verifying its effectiveness and adaptability in real scenarios. The improved RRT algorithm proposed effectively enhances the execution rate and environmental adaptability of robotic arms in path planning, providing a practical and feasible technical approach for the autonomous operation of robotic arms in the field of intelligent manufacturing.