Developed artificial intelligence (AI) technology, especially the breakthrough in reinforcement learning, Q-learning solves various decision and optimization problems owing to its simplicity and effectiveness. A simulation game platform based on Q-learning was designed in this study to simulate different offensive and defensive confrontation scenarios. A large number of experiments were conducted on the platform, and the effectiveness of the proposed improved Q-learning framework was verified paper. The improved algorithm converges to the optimal strategy faster than traditional Q-learning methods in offensive and defensive confrontation scenarios. It significantly improves the tank’s winning rate and combat efficiency.