Physical Plausibility-aware Trajectory Prediction via Locomotion Embodiment

This supplementary material covers details and additional results that could not be included in the main manuscript due to page limitations. First, Sec. A describes the implementation details and experimental setup. Section B presents additional results with EmLoco loss, including results on another real-world dataset, additional analyses, and more qualitative evaluation. Section C provides further evaluations of the LocoVal filter, including results with different HTP network baselines, detailed analyses, and visualizations. Lastly, Sec. D discusses failure cases and possible future directions.