【智远讲坛第19期】戴盟机器人首席技术官杜仪湃：视触觉与具身智能操作应用

接触发生后，受力、形变、材质、摩擦与滑移等核心物理交互信息无法仅凭视觉判断，缺少触觉反馈，强大的通用基础模型也难突破精细操作的瓶颈。本报告将围绕这一问题，从基于光学成像原理的单色光视触觉技术出发，介绍真实物理交互信息的采集，为模型学习真实接触与物理因果提供数据基础。

Once contact is established, key aspects of physical interaction—including contact forces, deformation, material properties, friction, and slip—cannot be reliably inferred from vision alone. Without tactile feedback, even highly capable general-purpose robotic foundation models remain limited in fine-grained manipulation. This talk addresses this challenge by introducing monochromatic vision-based tactile sensing based on optical imaging. It presents the acquisition of real-world physical interaction data, providing an empirical basis for models to understand contact interactions and reason about physical cause-and-effect relationships.

杜仪湃博士本科毕业于上海交通大学密西根学院，硕士毕业于瑞典皇家理工学院（全奖），博士毕业于香港科技大学机器人研究所，曾任美国普渡大学MARS实验室访问学者。其研究领域为基于视觉的触觉传感、机器人操作以及机器人学习。