In order to overcome the shortcomings of the basic genetic algorithm (GA) for solving dynamic path planning for mobile robot, an improved GA is proposed. Firstly, grid method is used to model the path planning for mobile robot. Then, the method for generating the initial population and selecting elite strategy is proposed, and an adaptive mutation probability is designed in order to improve the quality of algorithm solution. Meanwhile, in the process of planning, according to different types of robots colliding with the dynamic obstacles, the corresponding collision avoiding strategies are proposed by combining the global and local path planning. Simulation results show that the proposed algorithm is superior to the basic GA. It can effectively guide the robot in dynamic environment to realize the obstacle avoidance and get no touching optimal or suboptimal path.