您查看的活动已结束

Impact-aware humanoid control

Abstract /摘要

Almost all the robots fear impacts. Without accurate impact mechanics, robots typically cautiously approach rigid contacts at a near-zero velocity. Despite the human-like morphology, the humanoid robots can not apply human-like swift motion in a contact-rich environment. Thus, humanoid robots rarely, if not impossible, impress the general public and solve practical applications. Modeling and controlling the impact event is challenging. Impacts cause sudden changes in the velocities and torques.These state jumps can severely damage the robot hardware, enable un-deterministiccontact mode (e.g., sliding or sticking), and, most importantly, break the standing stability. We envision future humanoid robots not worrying about impacts and can assist humans through energetic physical contact. This talk will go through the latest impact mechanics modeling and control. In particular, the state-of-the-art usually focuses on planning periodic impacts without ensuring the hardware/control feasibility during the time interval before and after the impact event. We base our control design on the set of all the candidate frictional impulses and enable hardware-affordable dynamic motion. The on-purpose impact event becomes deterministic and undercontrolled.Preliminary experiments employing the Panda manipulator and the humanoid robot HRP-4 proved that impact awareness significantly improves the robot'smanipulation efficiency without breaking hardware or falling.

Short bio/个⼈简介:

Wang, Yuquan received his Bachelor's degree in Mechanical Engineering and Automation from Beijing Jiao Tong University in 2008, a M.S. degree in System

s, control and robotics, and a Ph.D. degree in Robotics and computer vision from the Robotics, Perception, and Learning (RPL) at the Royal Institute of Technology (KTH), Stockholm, Sweden, in 2010 and 2016 respectively. He worked as a researchassistant at the Eindhoven University of Technology in 2011; as a Post-doc researcher in the Sustainable Production Systems group at KTH between 2016 and 2018. Before he joined Maastricht University in September 2022, he served as a Post-doc researcher at CNRS-UM, LIRMM. His research focuses on modular and flexible control design that enables robust, safe, and easy-to-use robotic systems.

Advanced Humanoid Robotics – Biped walking control and locomotion planning

Abstract:

The human environment is tailor to way human moves around. As a result the robot that operates in such an environment has to walk. Bipedal walking is a task that’s very easy for humans, but very hard for robots, mainly because standing robot is underactuated and inherently unstable. During the motion planning and motion execution phase special attention must be given to contact forces acting on the robot’s feet. The contact between the feet and the ground has to be constantly maintained to ensure stable walking. In this lecture we will start from walking stability indicator ZMP, introduced more than 50 years ago and go up to state of the art walking stability indicators, capable of handling inclined surfaces and multiple contacts. We will take care about the underlying physical constraints and mention appropriate planning and control approaches that are built to fulfill them.

Speaker：

Milutin Nikolić received his M.Sc in Mechatronics and a Ph.D. degree in Robotics from the University of Novi Sad, Serbia in 2008 and 2015 respectively. He currently works as an Associate professor of Robotics at the University of Novi Sad, Faculty of Technical Sciences, Chair of Mechatronics, Robotics, and Automation. His field of research includes whole-body motion synthesis, walking pattern generation, contact stability criteria, multi-body system dynamics and robot manipulation and grasping. He spent 18 months at Nakamura-Yamamoto lab, the University of Tokyo in Japan, as a Project assistant professor working on human motion capturing and analysis. Nikolić was also working as a research associate at NTU Singapore, TU Technikum Wien, and UMIT, Hall in Tyrol. Nikolić has industry experience, working as a principal robotic arm manipulation engineer at the Aeolus robotics, where he developed novel approaches in arm calibration and whole-body motion planning. He also gave several guest lectures at well-established robotics laboratories, some of which are Nakamura & Yamamoto Lab University of Tokyo Japan, Biorobotics Laboratory, EPFL, Switzerland, Vislab at  Instituto Superior Técnico in Lisbon Portugal, Automation, Biocybernetics and Robotics Department at Jožef Stefan Institute….

​