The bidirectional rapidly random exploring tree star (BI-RRT*) algorithm has issues such as slow sampling speed, lengthy paths, and excessive turning points in global path planning. To solve these issues, this paper presents an improved BI-RRT* algorithm. Firstly, the Artificial Potential Field (APF) theory is integrated to enhance the efficiency and directionality of the RRT tree expansion. Secondly, the elliptical space sampling strategy is adopted to limit the sampling domain and accelerate the convergence process. Additionally, a node constraint strategy is employed to minimize excessive exploration of ineffective areas and eliminate useless tree nodes. Furthermore, a path optimization strategy based on the triangle inequality principle is introduced, to accelerate the convergence of the random tree to the best solution by straightening its curves. Comparative experiments demonstrate that the improved BI-RRT* algorithm achieves higher search efficiency, stronger purposiveness and smoother planning paths.