Abstract
Constructing action targets from measured robot motion is an established approach in imitation learning. Under interaction constraints, however, command-state discrepancy may reflect control demands that motion alone does not capture. We investigate when this information matters and how to exploit it. Across three real-robot tasks, task and phase analyses reveal larger supervision gaps under constrained interaction, while selective command retention provides evidence of locally useful command information. Building on these findings, we propose Command-State Discrepancy Weighting (CSDW), which accounts for robot response times and combines subsequent progress, persistent unmet demand, and demand changes into continuous weights for command supervision. The method requires no task-phase annotations or changes to policy architecture or inference. CSDW improves over uniform command supervision on constrained tasks, while methods perform similarly in the less constrained task. Project page: https://seen-e.github.io/CSDW/.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Li, P., Tao, Y., Zhang, E., Zhao, Z., Yue, C., Wang, H., Lv, L., Zhao, W., Chen, J., Liu, X., Huang, K., Luo, Y., & Liu, H. (2026). Beyond State-as-Action: Exploiting Command-State Discrepancy for Robot Imitation Learning. https://omanscience.com/en/articles/beyond-state-as-action-exploiting-command-state-discrepancy-for-robot-imitation-learning
MLA 9
Li, Peiyan, et al. "Beyond State-as-Action: Exploiting Command-State Discrepancy for Robot Imitation Learning." https://omanscience.com/en/articles/beyond-state-as-action-exploiting-command-state-discrepancy-for-robot-imitation-learning.
Chicago (author–date)
Li, Peiyan, Yueran Tao, Enhao Zhang, Zhixuan Zhao, Chenghao Yue, Hao Wang, Lei Lv, Wentao Zhao, Jiahao Chen, Xin Liu, Kangyao Huang, Yu Luo, and Huaping Liu. 2026. "Beyond State-as-Action: Exploiting Command-State Discrepancy for Robot Imitation Learning." https://omanscience.com/en/articles/beyond-state-as-action-exploiting-command-state-discrepancy-for-robot-imitation-learning.
Harvard
Li, P., Tao, Y., Zhang, E., Zhao, Z., Yue, C., Wang, H., Lv, L., Zhao, W., Chen, J., Liu, X., Huang, K., Luo, Y. and Liu, H. (2026) 'Beyond State-as-Action: Exploiting Command-State Discrepancy for Robot Imitation Learning', Available at: https://omanscience.com/en/articles/beyond-state-as-action-exploiting-command-state-discrepancy-for-robot-imitation-learning.
Vancouver
Li P, Tao Y, Zhang E, Zhao Z, Yue C, Wang H, et al. Beyond State-as-Action: Exploiting Command-State Discrepancy for Robot Imitation Learning. https://omanscience.com/en/articles/beyond-state-as-action-exploiting-command-state-discrepancy-for-robot-imitation-learning
IEEE
P. Li, Y. Tao, E. Zhang, Z. Zhao, C. Yue, H. Wang, L. Lv, W. Zhao, J. Chen, X. Liu, K. Huang, Y. Luo, and H. Liu, "Beyond State-as-Action: Exploiting Command-State Discrepancy for Robot Imitation Learning," https://omanscience.com/en/articles/beyond-state-as-action-exploiting-command-state-discrepancy-for-robot-imitation-learning.