Too Stiff, Too Strong, Too Smart: Evaluating Fundamental Problems with Motion Control Policies

dc.contributor.authorXie, Kaixiangen_US
dc.contributor.authorXu, Peien_US
dc.contributor.authorAndrews, Sheldonen_US
dc.contributor.authorZordan, Victor B.en_US
dc.contributor.authorKry, Paul G.en_US
dc.contributor.editorWang, Huaminen_US
dc.contributor.editorYe, Yutingen_US
dc.contributor.editorVictor Zordanen_US
dc.date.accessioned2023-10-16T12:32:58Z
dc.date.available2023-10-16T12:32:58Z
dc.date.issued2023
dc.description.abstractDeep reinforcement learning (DRL) methods have demonstrated impressive results for skilled motion synthesis of physically based characters, and while these methods perform well in terms of tracking reference motions or achieving complex tasks, several concerns arise when evaluating the naturalness of the motion. In this paper, we conduct a preliminary study of specific quantitative metrics for measuring the naturalness of motion produced by DRL control policies beyond their visual appearance. Namely, we propose to study the stiffness of the control policy, in anticipation that it will influence how the character behaves in the presence of external perturbation. Second, we establish two baselines for strength that allow evaluating the use of joint torques in comparison to human performance. Third, we propose the study of variability to reveal the unnatural precision of control policies and how they compare to real human motion. In sum, we aim to establish repeatable measures to assess the naturalness of control policies produced by DRL methods, and we present a set of comparisons from state-of-the-art systems. Finally, we propose simple modifications to improve realism on these axes.en_US
dc.description.number3
dc.description.sectionheadersPhysics-based Character Control
dc.description.seriesinformationProceedings of the ACM on Computer Graphics and Interactive Techniques
dc.description.volume6
dc.identifier.doi10.1145/3606935
dc.identifier.issn2577-6193
dc.identifier.urihttps://doi.org/10.1145/3606935
dc.identifier.urihttps://diglib.eg.org:443/handle/10.1145/3606935
dc.publisherACM Association for Computing Machineryen_US
dc.subjectcomputer animation, motion control, reinforcement learning
dc.subjectcomputer animation
dc.subjectmotion control
dc.subjectreinforcement learning
dc.titleToo Stiff, Too Strong, Too Smart: Evaluating Fundamental Problems with Motion Control Policiesen_US
Files