Human motion prediction improved with graph structured flow matching
Skeletons in Flow: Graph Structured Flow Matching for Human Motion Prediction
Computer Vision and Pattern RecognitionRobotics
Summary
Predicting how people move in the future is tricky because the body has many connected joints that move together over time. The authors introduce a new method called Graph Structured Flow Matching (GSFM) that predicts full future body motions by modeling the whole body's joint movements as a connected graph changing over time. Their method keeps bone lengths constant and respects how joints relate spatially and temporally. They tested GSFM on large motion datasets and also showed it works on different skeletons without extra training.
What this means in practice
- •For animation studios: Generate diverse and physically consistent character motions for game and film animation based on past movement data.$Commercial implications: Enables production of realistic animated characters with fewer manual adjustments, improving content creation efficiency and quality.
- •For robotics engineers: Improve humanoid robot motion planning by accurately predicting natural joint movements considering spatial and temporal constraints.
Authors
Yixuan Wang, Brandon C. Fallin, Warren E. Dixon
Abstract
Human motion prediction requires diverse future trajectories that remain consistent with observed motion and the articulated physical structure of the body. Skeletal constraints restrict individual poses, while coordinated motion depends on spatial interactions (between connected joints) and temporal interactions (between time instants). To facilitate human motion prediction in light of these constraints and interactions, we introduce Graph Structured Flow Matching (GSFM), which transports the complete future skeletal trajectory through a single conditional velocity field. The trajectory produces a spatiotemporal skeleton graph, and spatial and temporal attention couple its evolution according to skeletal relations and physical time offsets. Bone directions lie on unit spheres relative to a root joint, and tangent evolution preserves input bone lengths throughout generation. We train a learned velocity field through conditional flow matching along geodesic paths connecting random trajectories centered on the last-observed pose to recorded future trajectories. Experiments on the Archive of Motion capture As Surface Shapes (AMASS) dataset evaluate prediction accuracy, diversity calibration, and motion statistics. We demonstrate the contributions of spatial and temporal message passing in the developed architecture through an ablation study. GSFM models trained on AMASS also perform competitively on the Human3.6M skeleton without parameter updates or retraining, demonstrating applicability to an unseen skeletal structure.