Manuel Yves Galliker

1X Technologies

Thursday, April 04, 2024, 11:00 - 11:59

SR 01-012

In order to allow legged robots to be useful and benefit humanity they need to be able to robustly control a wide range of locomotion and manipulation behaviors.

State of the art solutions to this problem, encompassing closed-form solutions using simplified models, online optimization, or offline reinforcement learning, are fundamentally dependent on the ability to accurately model and simulate the dynamics of the problem. In this talk I will present our recent work on robot control and motion planning and provide an overview how it enables our Androids to collect data to perform autonomous loco-manipulation tasks. In particular the talk will highlight how different model structures, including template models, centroidal and whole-body dynamics, can be used and combined with terminal invariant sets in the context of Nonlinear Model Predictive Control to achieve a wide range of legged locomotion behaviors. 

Furthermore, we discuss arising challenges including the stiffness of the optimization problem, energy efficient locomotion, natural gaits and computational complexity for the presented approaches and beyond.

Bio:

Manuel Yves Galliker is the team lead for controls and embedded at 1X Technologies. He holds a B.Sc. and M.Sc. in Mechanical Engineering with a focus on Robotics, Systems and Controls from ETH Zurich. Manuel’s research interests range from mechantronics to reinforcement learning,optimal and data-driven controls with a focus on locomotion and loco-manipulation. During his academic tenure, he dedicated his efforts of his master thesis to researching online gait generation for bipedal robots using Whole-Body Dynamics in a Nonlinear Model Predictive Control approach, contributing at ETH’s Robotics Systems Lab and as a visiting researcher at Caltech’s AMBER lab. This work culminated in a publication presented at the IEEE Humanoids 2022 conference, where it was distinguished as a finalist for the Best Paper Award. In his current role at the Humanoid robot Scale-up 1X Technologies, he is leading the R&D efforts on controls and embedded for the new bipedal Android NEO. In particular the team is aiming to develop whole body control and planning algorithms, ranging from Centroidal MPC to Whole-Body MPC and Reinforcement Learning, to enable consecutively more and more general loco-manipulation behaviors.


Also online via Zoom:
https://uni-freiburg.zoom.us/j/62791737415?pwd=UDJnbkZlS3NkVm1TSVZLSWxHSktZZz09
Meeting ID: 627 9173 7415
Passcode: syscop2021