Proactive human-robot collaboration (HRC), as a cognition-centric approach, aims to reason the dynamic process of tasks for proactive robotic decision-making, which can be represented through the spatiotemporal evolution of non-paired relationships. Most existing works rely solely on vision-driven knowledge graph methods to reason the spatiotemporal evolution. However, the spatiotemporal evolution of non-paired relationships involves the interaction of multimodal information, and understanding such interactions requires robust analytical capabilities, which poses challenges for proactive robotic decision-making. This paper proposes an interactive semantics-enhanced vision-language model-driven spatiotemporal hypergraph reasoning method (VLSHR) to reveal the spatiotemporal evolution of non-paired relationships. First, to understand vision-language semantics, we fine-tuned a vision-language large language model (LLM) with interactive semantics. Furthermore, vision-language semantics need to be transformed into a hypergraph structure that can represent non-paired relationships. To reason the spatiotemporal evolution of non-pairwise relationships in HRC, we define temporal hyperedges, spatial hyperedges, and task hyperedges, coupling the affiliations of nodes with different types of hyperedges to construct a spatiotemporal hypergraph for HRC tasks. Then, a spatiotemporal hypergraph neural network is developed to reason the spatiotemporal evolution of non-pairwise relationships for proactive robotic decision-making. Finally, a case study on HRC assembly tasks demonstrates the effectiveness of the proposed method.

ABSTRACT Current AI animation generation methods excel in tight‐fitting clothing scenarios but struggle with deformation distortion and the gradual loss of wrinkles over extended simulations in loose‐fitting clothing. To address these issues, we propose a multi‐attribute‐aware Graph Network. This approach mitigates the gradual loss of wrinkles by dividing animation sequences into multiple stages based on motion categories, recognizing that identical body postures can cause different clothing deformations due to varying motion tendencies. In each stage, we first restore coarse, globally guided deformations based on the motion category, followed by enhancing detailed features. We observed that garments within the same sport category exhibit similar local wrinkles and that the degree of fit to the body varies significantly across different regions of the same garment. We introduce two specific clothing attributes: “looseness” and “deformity,” which relate to local wrinkles and have physical significance. A clothing attribute encoder perceives these attributes and constructs a clothing graph model to estimate detailed features. Our method effectively handles clothing deformations across various motion types, including extreme postures, with qualitative and quantitative analyses confirming its effectiveness.