Aviation ASI agent decision-making model based on causal emergence

Aiming at the problems such as decision delay, computational complexity, and insufficient policy interpretability faced by multi-agent dynamic games in unmanned air combat systems, this paper proposes an aviation ASI agent decision-making framework based on causal emergence theory. By integrating the Hawkes point process and the Granger causality discovery method, a formation game dynamics model is constructed to achieve the macroscopic state compression of high-dimensional situations and the extraction of key causal correlations. We design a hybrid equalization solution mechanism, and combine the QMIX algorithm and attention calibration technology to optimize the multi-agent collaboration strategy. The experiment is based on the Monte Carlo simulation and deduction system to verify the nonlinear influence of offensive and defensive strategy parameters on combat effectiveness: The offensive strategy gains an advantage through early firepower suppression, while the defending side significantly weakens the attack effect through the dynamic resource management mechanism. Although the unlimited ammunition strategy increases the firepower density, it leads to a decline in tactical efficiency due to the constraint of mobile energy consumption. The results show that resource constraints and elastic allocation are the dominant factors in winning or losing. The proposed framework provides theoretical support for the dynamic strategy switching and equilibrium optimization of air combat agents and lays the foundation for the design of autonomous decision-making systems in complex confrontation environments.