报告题目：Deformable Object Manipulation with Transformer-Based Model
报告人：李帅
时间：2025年11月7日下午3:00
地点：实验楼D1栋401室
报告简介：Deformable manipulation has drawn great attention in robotics, especially for medical applications. But handling deformable objects is not easy and their complex, nonlinear behaviors make accurate control a real challenge. To address this, we introduce an online iterative perception policy. This method does not require large-scale training. Instead, it iteratively perceives changes and adjusts control actions on the fly, enabling precise and efficient manipulation. We tested this method in both simulations and real-world settings, comparing it with several advanced control methods. The results show that the proposed method achieves better stability, robustness, and flexibility, regardless of material properties or starting conditions.
报告人简介：李教授现任芬兰奥卢大学全职教授，同时担任芬兰国家技术研究中心（VTT）兼职教授。其研究方向涵盖机器人安全操作、标定及多机器人协作。已发表同行评审论文300余篇，总被引量达25000次，H指数85（谷歌学术）。目前为英国工程技术学会会士、英国计算机学会会士、亚太人工智能协会会士。
