Social Perception and Formation Control for Human–Robot Co-Working Systems

The coexistence of humans and collaborative robots (cobots) has become an inevitable trend in modern industry. Previous research mainly focuses on how a cobot interacts with a single human and often struggles to interpret complex or crowded environments. To address these limitations, this paper presents an integrated social-perceptual framework for collaborative robots to deliver flexible behaviors in front of human or a group of humans. Unlike conventional APF approaches that focus solely on obstacle avoidance based on distance metrics, this integrated framework synergizes social perception cues and collaborative formation modeling into a unified potential-field formulation for practical industrial use. Initially, five task-oriented formations of cooperative workers that commonly occur in shared workspaces such as L for loading/handover, C for closed collaboration, T for transferring, V for vision-based supervision and S for surround/multi-person, are established through mathematical formulations. Subsequently, a social model for comfortable human-oriented interaction is introduced to ensure the reasonable collaboration without anxiety. Later, our strategy to establish the dynamic risk potential estimation (DRPE) is developed for distinguishing the formation zone owing to the interactive rules. Accordingly, the proposed scheme of gaze-attention for collaborative behavior (GA4CB) is introduced to determine the appropriate co-working interaction zones. To validate the feasibility, effectiveness, and application of our approach, both numerical simulations and experimental investigations are carried out. From these results of validations, it can be seen clearly that our method is appropriate in various contexts of industrial manufacturing.