path planning problem for soccer robots in the unknown environment of RoboCup competition is researched. This paper proposes a MOD-RRT soccer robot path planning algorithm, which solves the problem that the soccer robot has slow speed and poor effect in path planning. According to the principle of basic RRT algorithm, aiming at the shortcomings of strong randomness, slow convergence and poor path smoothing effect, optimization methods such as increasing the gravitational component at random sampling points, multi-step extended approximation target points and tailoring and path smoothing of redundant nodes are proposed. The simulation experiments in different obstacle maps show that the optimized planning path length is reduced by about 20%~30%,the number of sampling nodes is reduced by about 45%~65% compared with the basic rapidly-exploring random tree algorithm. Finally, the optimization algorithm is transplanted to the SimRobot simulation platform. The results show that the MOD-RRT algorithm has good real-time and robustness in the unknown environment, which can meet the path planning requirements of the robot on the competition.