ku酷游app入口,亚博app,皇冠hg666体育

讲座名称：Motion generation and transfer learning for robotics

讲座人：Sebastien Lengagne 副教授

讲座时间：5月25日19:00-20:00

讲座地点：雁塔校区主楼III区201报告厅

讲座人介绍：

塞巴斯蒂安?朗加涅自 2013 年起任职于法国克莱蒙费朗帕斯卡研究院，担任副教授。于2009 年在法国蒙彼利埃的机器人学、微观机械与微电子实验室（LIRMM） 获得机器人学博士学位，研究方向为机器人安全运动的运动规划与快速重规划。此后，他先后在日本筑波联合机器人实验室（JRL）、德国卡尔斯鲁厄理工ku酷游app入口,亚博app,皇冠hg666体育（KIT） 以及法国 LIRMM 实验室从事博士后研究，研究领域涵盖仿人机器人运动生成、多接触点运动、人体动作迁移以及足式机器人。其主要研究方向包括：仿人机器人与足式机器人的运动生成、区间分析与可靠优化、全身控制。近期研究重点为机器人领域的迁移学习。

讲座内容：

Generate complex robotic motions while providing guarantees on safety, stability, and performance is very important issue in Robotics. As robots become increasingly integrated into human environments, reliability becomes a critical issue. Robotic systems must therefore not only be efficient, but also predictable and trustworthy.Motion generation for humanoid and legged robots, especially in dynamic and multi-contact situations, is challenging task. The goal is to compute feasible whole-body motions while satisfying constraints related to balance, collision avoidance, friction limits, and actuator capabilities. To tackle these problems, interval analysis and set-based methods have been investigated.The central idea is to compute over sets instead of single values, making it possible to obtain mathematical guarantees over continuous trajectories and optimization processes. This framework allows the validation of constraints over the entire motion duration and provides a unified formulation for equality, inequality, and derivative constraints. These methods were validated on several robotic applications, including dynamic walking, obstacle crossing, multi-contact locomotion, and navigation in constrained environments.More recent work has focused on transfer learning in robotics. Analytical model-based methods provide strong guarantees but remain limited when dealing with highly complex motions.Conversely, learning-based approaches can generate sophisticated behaviors, but they require significant training time and computational resources. The objective of this research direction is therefore to combine the advantages of both approaches by designing learning frameworks capable of transferring skills between robots with different morphologies and different numbers of degrees of freedom. Overall, this work aims to bridge model-based robotics and machine learning in order to develop robotic systems that are both efficient and reliable, while preserving the guarantees required for safe interaction with humans.

主办单位：电子工程ku酷游app入口,亚博app,皇冠hg666体育

亚博app报名二维码：