V3I6P106

Adaptive Guidance Law Design for Air-to-Air Missiles via Reinforcement Learning

Kieu Tien Dung*¹

Abstract

This paper presents an adaptive guidance law design method for air-to-air missiles using reinforcement learning (RL). Conventional guidance laws, such as Proportional Navigation (PN), exhibit limitations when engaging highly maneuvering targets under uncertain combat conditions. To address this challenge, an RL-based guidance framework is proposed, enabling the missile to learn optimal interception strategies through continuous interaction with a simulated engagement environment.

The missile-target engagement is formulated as a Markov Decision Process (MDP), where the state space includes relative position, velocity, line-of-sight rate, and engagement geometry. The reinforcement learning agent generates guidance commands by maximizing cumulative rewards associated with interception accuracy, control efficiency, and engagement success. A deep neural network is employed to approximate the optimal policy and improve decision-making in nonlinear and dynamic scenarios.

Simulation results demonstrate that the proposed method achieves higher interception probability and lower miss distance than conventional proportional navigation guidance, particularly against maneuvering targets. Furthermore, the learned guidance policy exhibits strong adaptability to variations in target behavior and engagement conditions. These findings indicate that reinforcement learning offers a promising approach for developing intelligent and adaptive guidance laws for next-generation air-to-air missile systems.

Keywords:

Air-to-air missile, guidance law, reinforcement learning, deep reinforcement learning, target interception, adaptive guidance.