发表机构
Toyota Technological Institute（丰田工业大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出能量正则化模仿学习方法，通过可微分能量预测器将机械功作为正则项微调操作策略，在RLBench上降低2.1%平均功并提升成功率，实现无需显式动力学模型的节能操作。
AI 中文摘要
本文研究能量感知操作作为一个物理基础的学习问题。我们利用关节力矩和角位移定义了一个关节空间机械功代理，并训练了一个可微分的能量预测器，该预测器从机器人状态和动作中估计此功。该预测器将模拟器侧不可微分的物理量转化为可微分的正则化项，用于微调预训练的操作策略。我们在RLBench上使用RVT-2实例化了该框架，并评估了12个涉及物体接触、关节运动、放置、推和清扫的操作任务。所提出的微调将平均机械功从208.8J降低到204.4J（即降低2.1%），同时平均任务成功率也从86.2%略微提升至86.9%。这些结果表明，功感知的策略优化可以抑制物理上低效的运动，而无需显式的可微分动力学模型。
英文摘要
This paper studies energy-aware manipulation as a physically grounded learning problem. We define a joint-space mechanical-work proxy from joint torque and angular displacement, and train a differentiable energy predictor that estimates this work from robot states and actions. The predictor converts a non-differentiable simulator-side physical quantity into a differentiable regularizer for fine-tuning a pretrained manipulation policy. We instantiate the framework with RVT-2 on RLBench and evaluate 12 manipulation tasks involving object contact, articulated motion, placement, pushing, and sweeping. The proposed fine-tuning reduces the average mechanical work from 208.8J to 204.4J (i.e., 2.1% reduction), while the mean task success rate also increases slightly from 86.2% to 86.9%. These results show that work-aware policy optimization can suppress physically inefficient motion without requiring an explicit differentiable dynamics model.
CommentsECCV 2026 Workshop on Force-Grounded, Cross-View Articulated Manipulation
