In this paper, combined with dynamic collision preventation and sliding mode control, a firefly algorithm based on the framework of a cultural algorithm is proposed to solve the dynamic path planning for the soccer robot problem. The convergence of the algorithm is proved by the mathematical theorem. According to the different tasks of the robot soccer in the competition, two roles of attack and defense are analyzed and discussed respectively. When attacking, combined with the method of dynamic collision avoidance, the path is smoothed and corrected. When defending, the trajectory of the ball or opponent robot is tracked based on sliding mode control, and the parameters of sliding mode control are optimized using the CFA algorithm. Thus the speed and angular velocity of the robot are computed. The robot soccer games are used to test the new algorithm. The experimental results confirms that the CFA has effectiveness and efficiency, regardless of without risk of collision, or with risk of collision when encountering different obstacle robots. The effects of different parameters on the convergence performance are tested, such as path sampling points, population scale and evolutionary iteration number. Compared with the PSO, ACO and other famous evolutionary algorithms, it is verified that the proposed algorithm is easier to search the global optimal solution and has better convergence performance.