A Proportional Navigation–Inspired Reward Framework for Reinforcement Learning–Based Missile Guidance
2026 11th International Conference on Recent Advances in Air and Space Technologies (RAST), İstanbul, Türkiye, 13 - 15 Mayıs 2026, ss.1-6, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/rast69551.2026.11672446
- Basıldığı Şehir: İstanbul
- Basıldığı Ülke: Türkiye
- Sayfa Sayıları: ss.1-6
- Yıldız Teknik Üniversitesi Adresli: Evet
Özet
Proportional Navigation Guidance (PNG) continues to be used in missile guidance systems due to its ease of implementation. However, it tends to lose effectiveness when targets perform complex maneuvers or engagement conditions change quickly. These have positioned missile guidance as a compelling problem for reinforcement learning research. In this study, a two-dimensional pursuit-evasion simulation environment was created to establish the fundamental dynamics of a missile-target engagement. In this simulation, training of RL agents, which have a reward function inspired by PNG, was performed. The resulting models are evaluated across two maneuver settings: randomly generated target maneuvers and predefined scenarios encompassing diverse engagement geometries and maneuver types. The results demonstrated that PPO achieves the most successful interception performance, reaching a hit rate of 71.3% in the randomized maneuver test configuration and 97.8% in the scenario-based maneuver test configuration. These experimental findings indicate that, in model training with a PNG-based reward system, the PPO-based system can outperform conventional PNG under dynamic engagement conditions.