Robotic teleoperation in hazardous environments: a review of feedback architectures, stability, and learning-based adaptation

Abstract

Teleoperation is fundamental to robotic manipulation in hazardous environments where direct human presence is unsafe or impractical. However, existing research addressing these challenges is fragmented across multiple domains, hindering a coherent system-level understanding. Although robotic autonomy has progressed substantially, fully autonomous execution in unstructured and safety-critical settings remains challenging due to perception uncertainty, complex contact dynamics, and the requirement for human contextual judgment. As a result, hazardous environment teleoperation effectiveness is constrained by feedback architecture design, communication latency, stability¨Ctransparency trade-offs, and limited observability of remote interaction states. This review addresses three questions: how architectures influence stability and transparency, how delay and sensing limitations affect interaction quality, and how learning-based methods support authority allocation and adaptation. Unilateral, bilateral, and multilateral feedback architectures are analysed to elucidate their implications for stability, transparency, communication burden, and cooperative task execution. Latency compensation and passivity-based stabilisation strategies are reviewed to clarify how delayed communication influences energetic coupling and closed-loop stability. Recent advances in learning-based approaches to authority allocation, model mediation, and link-aware adaptation are analysed as emerging mechanisms for enhancing interaction quality. The synthesis indicates that teleoperation performance emerges from tightly coupled interactions among communication dynamics, stability constraints, perception fidelity, and human authority allocation, indicating that reliable operation requires integrated system-level co-design.