10月24日， 西湖大学智能无人系统实验室举办了一场题为 “ChatGPT meets Temporal Logics (TL): an introduction to TL-based robot motion planning” 的讲座。本次讲座由北京大学的孙志勇教授和荷兰埃因霍芬理工大学的 Sofie Haesaert 教授联合主讲，共同探讨了如何将自然语言命令转化为机器人可以理解和执行的任务，这是机器人技术领域的一大挑战。
图1 孙教授和讲座现场师生
孙志勇教授首先介绍了其团队的研究成果，即基于大语言模型（LLM）的机器人规划控制器。该规划器使用信号时间逻辑（STL）作为自然语言和具体任务之间的桥梁，并加入了反馈与语义检查模块。孙教授提出的 ChatGPT+TL 机器人规划器能够精确地将命令转化为机器人可以执行的具体动作，极大地提高了任务执行的准确性和效率。
图2 Sofie教授和讲座现场师生
在讲座的第二部分，Sofie 教授进一步深入讲解了时间逻辑（TL）作为机器人任务规范的关键工具，并展示了其在机器人运动规划中的几个关键特性和应用潜力。时间逻辑不仅为机器人提供了一种精确的任务描述方式，还为机器人的自主决策和行为提供了强大的支持。
本次讲座吸引了西湖大学对大语言模型感兴趣的研究人员的广泛关注。参会者在讲座后与两位教授进行了问答和深入的交流。西湖大学智能无人系统实验室对孙志勇教授和Sofie Haesaert教授的到访表示感谢，并期待未来有更多的学术交流与合作机会。
讲座预告信息如下
时间：2024年10月24日（周四）10:30-12:00
Time: 10:30-12:00, Thursday, October 24, 2024
地点：云谷校区E10-212
Venue: E10-212, Yungu Campus
主持人: 工学院 赵世钰 博士
Host: Dr. Shiyu Zhao, School of Engineering
语言：英语
Language: English
主讲人简介/Biography:
Zhiyong Sun received the Ph.D. degree from The Australian National University (ANU) in February 2017. He worked as Research Fellow/Lecturer at ANU of Australia, then Postdoc Fellow at Lund University of Sweden, then Assistant Professor at Eindhoven University of Technology (TU Eindhoven) of Netherlands. He joined Peking University in the summer 2024. He has won the Springer Best PhD Thesis Award, and several best paper and student paper awards from CDC, AuCC, ICRA, CCTA and ICCA. His research interests include multi-robotic systems, autonomous motion planning, distributed control and optimization.
Sofie Haesaert received the Ph.D. degree in 2017 from TU Eindhoven and she did a postdoc at California Institute of Technology (Caltech). In 2017, she won the best PhD Thesis Award of the Dutch Institute of Systems and Control. She started as assistant professor at TU Eindhoven in 2018. Her research work specializes in the verifiable design of cyber-physical systems using data-driven modelling, control engineering, and formal verification.
讲座摘要/Abstract:
In this talk we introduce temporal logics (TL) for robotic task and planning, and propose a novel ChatGPT+TL robot motion planner. In the first part, we present a LLM-based motion planner with signal temporal logic (STL) specifications serving as a bridge between natural language (NL) commands and specific task objectives. The rigorous and abstract nature of formal specifications allows the planner to generate high-quality and highly consistent paths to guide robot's motion and control. In the second part, we elaborate temporal logics as one key tool in robot task specification, and present several key features and application potentials for robot motion planning.
西湖大学官网讲座信息链接：ChatGPT meets Temporal Logics (TL): an introduction to TL-based robot motion planning
