A data-driven learning guidance strategy for target interception with terminal time constraint

This paper proposes a data-driven learning guidance strategy to tackle the target interception problem with terminal time constraints. The developed guidance algorithm employs the prediction-correction concept to derive commands effectively. We first develop a two-branch neural network to predict time-to-go, assuming there is no constraints on target’s maneuvers. Using the predicted time-to-go, we then propose a biased proportional navigation guidance (PNG), where the the biased command is computed by a deep reinforcement learning (DRL) network, to correct the impact time. Results from intensive Monte-Carlo simulations validates the efficacy of the method.