Human–Robot Interaction in Physical Robot Tasks: Learning, Adaptation, and Safety

Physically embodied robots are increasingly expected to perform tasks alongside people and to interact with objects, tools, and complex environments. In such settings, task requirements, environmental conditions, and human needs may change during operation, making fixed autonomous behavior insufficient. Human–robot interaction can provide demonstrations, instructions, feedback, and shared-control inputs that support more capable and responsive robot behavior. Recent advances in robot learning, multimodal interaction, adaptive control, and safety-aware planning create new opportunities to integrate these inputs into physical robot tasks. However, major challenges remain in translating human interaction into effective task learning, online adaptation, and safe execution while preserving usability, appropriate autonomy, and reliable performance in real-world settings.This Research Topic aims to advance human–robot interaction for the learning, adaptation, and safe execution of physical robot tasks. Here, physical robot tasks are understood broadly as embodied tasks in which robots interact with people, objects, tools, or physical environments. Contributions may investigate how robots acquire task knowledge from demonstrations, language, preferences, corrections, or multimodal feedback; how they adapt to different users, task conditions, and environments; and how safety, transparency, and controllability can be maintained during execution. Rather than addressing human-robot interaction in general, the Topic seeks to connect robotics, artificial intelligence, control, and human factors through experimentally grounded systems performing useful tasks in realistic environments.This Research Topic welcomes Original Research, Review, Perspective, Methods, Data Report, and Technology and Code articles on human-robot interaction in physical robot tasks. Relevant themes include learning from demonstration and human feedback; interactive, imitation, preference-based, and reinforcement learning; language-based and multimodal instruction; human-guided planning and control, including corrective or assistive inputs during execution; adaptation to users, tasks, and environments; safe physical interaction; uncertainty and failure communication; explainable robot behavior supporting task execution; trust, workload, and usability evaluation; and real-world system validation. Applications are especially encouraged in industrial and collaborative manipulation, assistive and service robotics, medical robotics, field and mobile manipulation, and humanoid robotics. Contributions should explicitly show how human interaction changes the learning, adaptation, control, safety, or evaluation of a physically embodied robot task. Topic editors: Kuanqi Cai, Wenzheng Chi, Weinan Chen, Chao Tang Source: https://www.frontiersin.org/research-topics/86654/human-robot-interaction-in-physical-robot-tasks-learning-adaptation-and-safety