Strategy optimization for UAV adversarial game based on counterfactual regret minimization

With the increasing application of unmanned aerial vehicle (UAV) technology, there is a growing demand for strategy optimization in adversarial scenarios. In particular, characteristics such as UAV swarm operations, dynamic task allocation, and cross-domain cooperation have expanded the strategy space from traditional low-dimensional static models to high-dimensional dynamic environments. Although traditional game theory possesses a well-established theoretical system for static games, it still faces challenges such as high computational complexity and poor adaptability when dealing with decision-making in high-dimensional and dynamic contexts. The counterfactual regret minimization (CFR) algorithm, which offers strict convergence guarantees to the Nash equilibrium, provides an effective solution for dynamic games with imperfect information. This study applies the CFR algorithm to solve the UAV adversarial game. Addressing the characteristics of incomplete information, unknown opponent strategies, and dynamic evolution, a strategy optimization model based on dynamic game theory is constructed. Through the discretization of the state space and the design of an environmental reward function, effective dimensionality reduction and efficient strategy updates in high-dimensional state spaces are achieved. Simulation results demonstrate that our proposed approach exhibits high policy optimization efficiency, excellent environmental adaptability, and strong strategic stability. These findings offer a feasible solution for an autonomous UAV adversarial game, with broad application prospects in areas such as military defense and civilian cooperation.