06.1 - Flight Dynamics and Control (Control & Modelling)

OFFLINE REINFORCEMENT LEARNING FOR INFRARED IMAGE-BASED MISSILE GUIDANCE AND CONTROL

S. Kim¹, J. Shin¹, H.-G. Kim, Konkuk University, South Korea; ¹Chungbuk National University, South Korea

An offline reinforcement learning framework for missile guidance and control is proposed using expert data generated by a sliding mode guidance law. TD3+BC is adopted to mitigate distributional shift and Q-value underestimation by constraining the policy to expert behavior. Simulations demonstrate stable interception and a 1.2% higher success rate than the baseline.

View Paper

DOI: 10.71945/icas2026_0858