Research on Urban Combat Tactical-Level Action Prediction Model Based on Dynamic Force Fields and Deep Reinforcement Learning

In response to the complex dynamic situation and the characteristics of enemy-friend interaction in urban combat environments, this paper proposes a tactical-level action prediction model based on dynamic force fields and deep reinforcement learning. By constructing dynamic force fields to quantify the threat and tactical opportunities in the battlefield, combining multi-agent game modeling to achieve dynamic analysis of enemy-friend interaction behaviors, and using deep reinforcement learning (Deep Q-Learning) to optimize our side’s action strategies. Simulation experiments are conducted based on typical urban street scenes, covering multiple conditions such as complex terrain, dynamic threats, and resource constraints. The experimental results show that the model proposed in this paper significantly outperforms traditional algorithms in key indicators such as action prediction accuracy, path planning quality, and real-time response efficiency, providing effective support for urban combat tactical-level decision-making.