To address the shortcomings of high sampling blindness, excessive number of redundant points, and high spatial sampling repetition rate in the traditional Rapidly-exploring Random Trees (RRT) algorithm in mobile robot path planning, an improved RRT algorithm is proposed. Firstly, a dynamic region sampling method is implemented to limit the scope of new node generation, thereby mitigating the randomness associated with node production. Then, a dynamic step length method is proposed to minimize the number of iterations required by the algorithm while reducing the time and length of the path-planning process. Finally, the path optimization strategy is designed to prune redundant nodes from the generated path, thereby refining and optimizing the final path. Comparison experiments are conducted through simulations with the RRT and gaussian sampling Goal-Bias RRT (GB-RRT) algorithms in large obstacles, narrow obstacles, and maze environments. The experimental results confirm the superiority of the improved RRT algorithm in the number of sampling nodes, path length, and sampling time.