Aiming at 3 kinds of weapons involved in island and reef garrison combat process, namely sea weapons, land weapons and air weapons, a model is established according to the island and reef garrison combat process, and a method of dynamic action space is proposed. 3 kinds of influencing factors are set, including enemy weapons and equipment, preset positions, and defensive points, and different reinforcement learning algorithms based on value function are used for testing.Through the test, the best position of each weapon and equipment can be obtained and whether the preset position is reasonable or not can be judged, and the comparison shows that the algorithms have their own advantages and disadvantages, and the suitable environments are different. The results show that the method can be applied to different environments, reduce the time and space overhead, improve the efficiency of island and reef garrison decision-making, and help to improve the strategy.