Aiming at the shortcomings of traditional A* algorithm planning such as long path calculation time, multiple search nodes, and multiple turning points, an improved A* algorithm based on weight coefficient optimization is proposed. By comparing Euclidean distance, Chebyshev distance, and Manhattan distance, it is shown that Manhattan distance has shorter search time and fewer turning points, so Manhattan distance is selected as the estimated cost function; By improving the weight coefficients of the actual cost function and the estimated cost function, the disadvantages of traditional A* algorithm such as long calculation time and multiple search nodes are solved; The resulting path is smoothed through a Bezier curve, resulting in a smooth path. Experiments show that the improved A* algorithm has short computing time, fewer search nodes, fewer turning points, and smooth trajectory.