Abstract: In recent years, path planning techniques for mobile robots have become increasingly important as their range of applications grows. Aiming at the problems of many search nodes, long optimal path length, many path bends in traditional A_star algorithm, an improved A_star global path planning algorithm is proposed. It improved the heuristic function and the search neighborhood method, and the dynamic weight processing and Bezier curve smoothing were carried out, realizing the effect of reducing the number of search nodes, shortening the path length, reducing the actual movement time of the mobile robot and reducing the number of path bends. The experimental validation was carried out based on the omnidirectional mobile robot experimental platform. The results conclude that the improved algorithm reduced the number of search nodes by 6.3%, the path length by 2.2%, the actual movement time of the mobile robot by 9.1% and the number of path bends by 44.4%. The improved algorithm meets more requirements of actual movement of mobile robot, proving the efficiency of the improved algorithm.