Abstract: To address the issues of excessive sampling randomness, low search efficiency, and zigzagging in traditional rapidly-exploring random tree (RRT) algorithms for robot path planning, we propose an improved RRT algorithm with a sampling area restriction strategy (SAR-RRT). First, to mitigate excessive randomness, we enhance the target orientation of the random tree by introducing a target bias strategy. And we constrain sampling using a spherical sampling region and an angular restriction strategy, which limits exploration in unnecessary spatial regions. Second, to improve search efficiency, we adaptively optimize the node expansion of the random tree. A multi-step expansion strategy is employed to maximize the use of environmental and obstacleinformation, while a greedy approach accelerates tree convergence, reducing path generation time. Finally, we apply secondary optimization to the initial planned path. After removing redundant points, we smooth the path using a cubic B-spline curve, improving overall path quality. Experimental results demonstrate that the SAR-RRT algorithm completes path-planning tasks in 2D and 3D scenarios. Compared with the traditional RRT algorithm, the proposed method reduces path length by 27. 73%, planning time by 85. 25%, and sampling points by 87. 19% while generating a smoother path.