GAMES Webinar 2026 – 415期(从世界表征到动力执行:具身操作中的物理智能)|杨理欣(上海交通大学), 窦志扬(麻省理工学院)

GAMES Webinar 2026 – 413期(三维内容的理解与生成)

报告嘉宾:杨理欣 上海交通大学

报告时间:2026年09月17日 晚上20:00-20:30(北京时间)

报告题目:

面向机器人操作的4D世界-动作表征

报告摘要:

本报告将探讨:“机器人如何理解动态世界的改变并予以动作生成”,以 3D Flow 为核心线索,梳理我们在时序流策略学习、追踪驱动的动作蒸馏、3DFlow+VLA/WAM架构上的系列探索。

讲者简介:

杨理欣,上海交通大学人工智能学院,助理研究员,博士生导师,卢策吾教授团队。上海创智学院具身标杆项目课题导师,主持国自然青年基金(C类)、上海市启明星-杨帆计划,研究领域为3D行为理解与具身智能;于2023年获得上海交通大学博士学位。他在人工智能与机器人领域知名期刊会议(如IEEE TPAMI, CVPR, NeurIPS, ICRA等)发表研究成果20余篇,获吴文俊自然科学奖一等奖项目“行为视觉理解”。

讲者主页:https://lixiny.github.io

报告嘉宾:窦志扬 麻省理工学院

报告时间:2026年09月17日 晚上20:30-21:00(北京时间)

报告题目:

NeuralActuator: Neural Actuation Modeling for Robot Dynamics and External Force Perception

报告摘要:

Differentiable simulators have advanced policy learning and model-based control across robotic tasks. Yet actuator dynamics remain underexplored and can be a major source of sim-to-real error, particularly on low-cost platforms, where the linear current-to-joint-torque approximation τ=KI becomes unreliable because of friction, hysteresis, backlash, and thermal effects. Accurate actuator models can also support force perception and integrated force/position control. We present NeuralActuator, which jointly predicts (i) a torque surrogate for trajectory propagation on low-cost servo platforms, (ii) external forces with a contact-probability gate for sensorless force perception, and (iii) a motor-condition score for a supervised joint, distinguishing normal from mechanically restricted operation. A twin-arm teleoperation system records robot states and actuator telemetry alongside external-force labels, yielding the Neural Actuation Dataset (NAD). The torque-surrogate head is trained through differentiable simulation from pose trajectories without ground-truth joint-torque measurements. A Transformer captures temporal dependencies while enabling real-time inference. We validate NeuralActuator on a 5-DoF OpenManipulator-X, a 6-DoF SO-101 from LeRobot, and a 7-DoF Franka Emika Panda, spanning three actuator families and costs from approximately $500 to more than $30,000. The low-cost platforms support physically plausible dynamics and force evaluation, while the offline Franka experiment provides a payload-force-estimation benchmark. We also demonstrate motor-condition estimation and improved behavior-cloning performance using NeuralActuator as a pretrained module. We release the dataset, code, and hardware configurations on the project page: https://frank-zy-dou.github.io/projects/NeuralActuator/index.html.

讲者简介:

窦志扬,麻省理工学院计算机科学与人工智能实验室(MIT CSAIL)计算机科学博士生,师从 Wojciech Matusik 教授,隶属计算设计与制造组和计算机图形学组。主要研究方向为物理智能、计算机图形学与具身智能,聚焦物理仿真、机器人动力学与控制、几何计算及人体运动与交互等。相关研究成果曾获 RSS 2026 杰出系统论文奖、SIGGRAPH 2023 最佳论文奖,并获 Computer Graphics Forum 年度高浏览论文和高被引论文荣誉。

讲者主页:https://people.csail.mit.edu/frankzydou/


主持人简介:

马月昕,上海科技大学长聘副教授、博士生导师,博士毕业于香港大学。主要研究方向为三维视觉、具身智能、自动驾驶。共发表相关领域顶会或顶刊论文100余篇,其中一作与通讯论文50余篇,包括TPAMI、CVPR、ICCV、SIGGRAPH等,谷歌学术引用近1万次。参与指导的论文获MICCAI 2024唯一最佳论文奖,ACM MM 2024最佳论文候选。担任TVCG、Visual Intelligence、RA-L编委,曾获上海市海外高层次人才、China 3DV 2025年度优秀青年学者、入选全球前2%顶尖科学家榜单等。

You may also like...