Title: “Towards optimal and learning control for gait free legged locomotion.”
A key challenge in legged robotics is to find suitable non-stereotypic behaviors (i.e. plans and stabilizing controllers) for a large range of different situations. Due to their non-trivial kinematics, multi-body dynamics and multi-contact physics, the dynamics of a legged robot changes dramatically in different configurations and from one moment to the next. The most pronounced of these effects is due to establishing and breaking contacts, which is a fundamental requirement for locomotion but also for manipulation. Yet planning and controlling of extended motions involving sophisticated multi-contact sequences poses a hitherto unsatisfactorily addressed challenge. I will discuss our work towards addressing the problem of finding optimal behaviors for robotic locomotion and manipulation problems via model-based optimal control and model-free learning algorithms. I will discuss the advantages and issues of a range of algorithms that promise useful in addressing this challenge. I hypothesize that the key to success is a smart combination of several methods and I will show some preliminary ideas of such meta algorithms.
Diego is a post-doctoral resarcher at the Agile and Dexterous Robotics Lab (ADRL) of the Institute of Robotics and Intelligent Systems ETH in Zurich, Switzerland. He is interested in the analysis and synthesis of challenging motor skills in complex robots. He is especially interested in robotic behaviors that occur when fast motion and body coordination are required. My goal is to develop control strategies that will allow robots to move as naturally, gracefully and efficiently as animals do.
Jonas Buchli is a SNF Assistant Professor for Agile and Dexterous Robotics at ETH Zurich since May 2012. He holds a Diploma in Electrical Engineering from ETH Zurich (2003) and a PhD from EPF Lausanne (2007). His thesis work is about adaptive frequency oscillators and modeling and design of robotic locomotion pattern generators. At EPFL he organized the 2006 Latsis Conference on ‘Dynamical principles for neuroscience and intelligent biomimetic devices’. From 2007 to 2010 he was a Post-Doc at the Computational Learning and Motor Control Lab at the University of Southern California, where he was the team leader of the USC Team for the DARPA Learning Locomotion challenge. From 2010-12 he was a Team Leader at the Advanced Robotics Department of the Italian Institute of Technology in Genova. Jonas Buchli has received a Prospective and an Advanced Researcher Fellowship from the Swiss National Science Foundation (SNF). In 2012 he received a SNF Professorship Award from the Swiss National Science Foundation. He has contributed to research in diverse fields such dynamical systems approaches to motion generation and control, the theory of coupled oscillators, planning and control of dynamic locomotion, machine learning, whole body control, whole body force and impedance control and modeling of human motor control. He was involved in the development of robotic platforms as well as software engineering projects for robotic control software. His current research interests include model based control of legged robotic and human locomotion and manipulation, machine learning and adaptive control, and dynamic and versatile service and field robots.