Artificial potential field approach is a common used method in robot path planning. But the composition of force is difficult to control when there are too many obstacles, the robot often failed to make path planning for there is trap of potential field. Enemy and friend condition in robot soccer system are complex, and also this is a dynamic area, and it is very important in real time computing. Based on the real time requests of the platform, this article will doing research on attention mechanism, which used on the robot path planning. In this modified method, physical characters are converted to mathematic expressions and integrated with traditional potential field functions. Based on the modified potential field method, change the composition of forces to gravitation force, and a robot soccer path planning is established, and its efficiency is proved by simulation experimentation.