V3I3P59

Deep Learning-Based Guidance Law for Intercepting Highly Maneuvering Targets

Tran Duc Khanh Hung^1*

Abstract

This paper proposes a deep learning-based guidance law for missile interception of highly maneuvering targets under uncertain and nonlinear engagement conditions. Traditional guidance laws such as Proportional Navigation (PN) are limited in handling complex target maneuvers and sensor noise. To address these limitations, a neural network-based guidance framework is developed to approximate the optimal control policy directly from engagement states. The proposed method is trained using a combination of supervised learning and reinforcement learning in a high-fidelity simulation environment. Numerical results demonstrate that the proposed approach significantly reduces miss distance and improves robustness compared to conventional PN guidance, especially against evasive targets.

Keywords:

Missile Guidance, Deep Learning, Reinforcement Learning, Proportional Navigation, Target Interception