Neurodynamics-Based Attack-Defense Guidance of Autonomous Surface Vehicles Against Multiple Attackers for Domain Protection

This article investigates an adversarial domain protection scenario involving multiple defending underactuated autonomous surface vehicles (ASVs) against an equal number of attacking ASVs in an obstructive environment. The defending ASVs can only obtain the positions of the attacking ASVs susceptible of stochastic measurement noises. A three-layer attack-defense guidance structure is proposed such that the defending ASVs are able to intercept the attacking ASVs with collision-free behaviors. In the first layer, an auction theory is utilized to achieve the attack-defense matching that efficiently assigns multiple defending ASVs to corresponding attacking ASVs in a distributed manner. In the second layer, an adaptive attack-defense guidance method is proposed, enabling each underactuated defending ASV to implement the appropriate guidance law aligned with the optimal interception point. In the last layer, a quadratic programming problem is formulated using speed-heading-constraint control barrier functions, and a neurodynamics optimization approach is employed to achieve real-time resolution of the optimal attack-defense guidance signal. The stability and safety analyses show that the system error is practically stochastic input-to-state stable, and the system is guaranteed to be safe. Experiment results demonstrate the effectiveness of the proposed neurodynamics-based attack-defense guidance method for domain protection.