Target-Aware Safety-Residual Reinforcement Learning for Cooperative Multi-UAV Pursuit in Complex Environments
Machines · 29.06.2026 · 14 (7) · 733
DOI: 10.3390/machines14070733 ↗Kurz gefasst
Fachartikel in der Zeitschrift Machines, online erschienen am 29.06.2026. Verfasst von Shun Li und 6 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Liaoning Provincial Science and Technology Plan Joint Programme.
Abstract (Originalsprache)
Multi-UAV cooperative persistent tracking in complex obstacle environments requires agents to approach dynamic targets while ensuring obstacle avoidance and flight safety; however, standard multi-agent reinforcement learning (MARL) methods typically rely on a single policy to implicitly handle both objectives, making it difficult to balance task performance and risk control. To address this issue, this paper proposes a Target-Aware Safety-Residual Pursuit Reinforcement Learning (TASRP) framework for constrained three-dimensional environments. A continuous-control 3D tracking environment is constructed in IsaacLab, where two multirotor UAVs cooperatively track a dynamic target under random, target-blocking, and gate-like obstacle layouts, boundary constraints, and inter-agent collision risks, with each UAV producing a four-dimensional action composed of normalized thrust and body-frame torques. TASRP adopts a dual-head residual policy in which a pursuit branch generates nominal actions, and a safety branch predicts corrective residuals, together with a risk-aware gating mechanism, a target-guided teacher for obstacle detouring, and a dual-critic safety-constrained optimization scheme. Under clean observations, TASRP achieves task success rates of 75–79%, obstacle crash rates of 13–15%, and boundary crash rates of 1–2% across three representative scenarios. Under noisy observations, TASRP achieves 72.1% task success, 20.3% obstacle crash, and 2.8% boundary crash, outperforming MAPPO (61.2%, 61.2%, 5.6%) and HAPPO (58.1%, 73.5%, 4.1%). These results indicate that explicitly decoupling target-oriented control and safety correction enables a more effective and robust performance–safety trade-off under both clean and moderately noisy observations.
Förderung
- Liaoning Provincial Science and Technology Plan Joint Programme
Zitieren
APA
Li, S., Yu, B., Liu, D., Gao, D., He, P., Chen, G., & Xu, L. (2026). Target-Aware Safety-Residual Reinforcement Learning for Cooperative Multi-UAV Pursuit in Complex Environments. Machines, 14(7), 733. https://doi.org/10.3390/machines14070733
Vancouver
Li S, Yu B, Liu D, Gao D, He P, Chen G, et al. Target-Aware Safety-Residual Reinforcement Learning for Cooperative Multi-UAV Pursuit in Complex Environments. Machines. 2026;14(7):733. doi:10.3390/machines14070733
BibTeX
@article{Li2026,
author = {Li, Shun and Yu, Bo and Liu, Dongying and Gao, Dayu and He, Peizheng and Chen, Gongbo and Xu, Lin},
title = {Target-Aware Safety-Residual Reinforcement Learning for Cooperative Multi-UAV Pursuit in Complex Environments},
journal = {Machines},
year = {2026},
volume = {14},
number = {7},
pages = {733},
doi = {10.3390/machines14070733}
}
EUVIMED