Dual-target domain attack-defense game based on intent recognition and task allocation

This paper investigates a two-versus-two reach-avoid differential game in a two-dimensional domain with dual targets, aiming to determine the defender’s optimal strategy to intercept attackers before they reach their respective targets. To this end, a dual-barrier structure is constructed based on geometric principles, effectively defining the defender’s advantageous region and establishing a new scenario for subsequent research. Furthermore, an RNN-LSTM neural network is adopted to recognize the attackers’ intentions in real time. Based on the inferred intent, the paper proposes an optimized defensive strategy that allows the defender to anticipate the attacker's target and respond accordingly. A matching matrix is then constructed to assign defense tasks among multiple defenders, leading to a more coordinated and efficient interception strategy. Finally, simulation experiments are conducted to verify the effectiveness of the proposed approach. The results demonstrate that the strategy achieves high performance in preventing successful intrusions and provides a foundation for future applications in large-scale defense scenarios.