A design method of differential game guidance law based on Dubins path and neural network

This study proposes a geometric solution to the norm differential game design problem in target-attacker-defender (TAD) engagements, addressing key limitations of conventional zero-effort-miss approaches. By leveraging the geometric analogy between guidance-law-generated trajectories and Dubins paths, we reformulate the derivation of zero-effort-miss-based guidance laws as a Nash equilibrium optimisation problem, with optimal strategies determined through reachable set analysis of Dubins path frontier. The resulting model is a non-convex optimisation problem, which prevents the derivation of traditional state-feedback control laws. To overcome this limitation and enable real-time implementation, we develop a custom back propagation neural network, enhanced with a relaxation factor method for output filtering, a Holt linear trend model for outlier compensation and a saturation function for oscillation suppression. Extensive simulations demonstrate that the proposed framework significantly outperforms baseline methods. These results validate the effectiveness and robustness of our approach for high-performance TAD applications.