Explicit-Reference-Governor Based Guidance Law Design Using Neural-Network Optimization

This paper studies the problem of missile intercepting against unknown maneuvering targets with a terminal impact angle under control constraints. Leveraging on the technique of explicit-reference-governor, control restrictions are transferred into upper bounds of the Lyapunov function, which follow the regulation of the derivative of applied reference for the constraint satisfaction. Our approach is further transferred into a convex program subject to nonlinear inequality constraints. Implementing a recurrent neural network for convex programs, the online solutions in the receding horizon manner provide upper bounds for the reference regulation and subsequently decisions of control in real time.