To address issues in traditional path planning algorithms such as blind search, node redundancy, inexecutable turns, and close adherence to obstacles when performing global path planning in complex environments, a comprehensive improved path planning algorithm based on Rapidly-exploring Random Tree (RRT) is proposed. First, obstacle expansion is performed. Then, a target attraction strategy and a dynamic step extension mechanism are introduced to accelerate the initial path search efficiency. At the same time, according to the vehicle kinematic model, turning angle constraints are added to the planned path to ensure executability. After obtaining the initial path, ellipse sampling and smart pruning strategies are applied until the iteration ends or the cost value converges. Experimental results show that this algorithm significantly improves planning time, path length, and the number of nodes.