Project

Back to overview

Learning Robust Control for Autonomous Robots

English title Learning Robust Control for Autonomous Robots
Applicant Buchli Jonas
Number 166163
Funding scheme SNSF Professorships
Research institution Institut für Robotik und Intelligente Systeme (IRIS) ETH Zürich
Institution of higher education ETH Zurich - ETHZ
Main discipline Other disciplines of Engineering Sciences
Start/End 01.09.2016 - 31.08.2018
Approved amount 586'576.00
Show all

All Disciplines (3)

Discipline
Other disciplines of Engineering Sciences
Mechanical Engineering
Electrical Engineering

Keywords (7)

rescue robots; locomotion; Autonomous Robots; learning control; manipulation; service robots; control and planning

Lay Summary (German)

Lead
Trotz gewaltiger Fortschritten in der autonomen Robotik sind komplexe robotische Systeme in vielen Bereichen nach wie vor nicht nutzbar. So zum Beispiel in der Rettungsrobotik, wo es wichtig ist das die Maschinen robust, sicher und mit einer gewissen Intelligenz agieren können um teils unabhängig eines überwachenden Operators eine Aufgabe auszuführen.
Lay summary

Inhalt und Ziel des Forschungsprojekts

Dieses Projekt hat das Ziel eine Regel und Planer Architektur für autonome Roboter, im speziellen Roboter mit Armen und Beinen zu entwickeln. Insbesondere soll dir Architektur und die dafür entwickelten Algorithmen robuste und selbstlernende Regler ermöglichen. Ein besonderes Augenmerk wird auf das robuste Verhalten der Regler und Planer gelegt.

Wissenschaftlicher und gesellschaftlicher Kontext des Forschungsprojekts

Mit den Resultaten dieses Projekt soll es ermöglicht werden dass komplexe Robotische Systeme, z.b. Roboter mit Armen und Beinen o. autonome Bauroboter die Bewegungsabläufe für komplexe Aufgaben zu einem guten Teil selbständig Planen und ausführen können. Damit soll der Einsatz solcher Maschinen in vielen wichtigen Bereichen der Gesellschaft ermöglicht werden (z.b. Rettungsrobotik).

Direct link to Lay Summary Last update: 17.09.2016

Responsible applicant and co-applicants

Employees

Publications

Publication
Automatic differentiation of rigid body dynamics for optimal control and estimation
Giftthaler Markus, Neunert Michael, Stäuble Markus, Frigerio Marco, Semini Claudio, Buchli Jonas (2017), Automatic differentiation of rigid body dynamics for optimal control and estimation, in Advanced Robotics, 1-13.
Fast Trajectory Optimization for Legged Robots Using Vertex-Based ZMP Constraints
Winkler Alexander W., Farshidian Farbod, Pardo Diego, Neunert Michael, Buchli Jonas (2017), Fast Trajectory Optimization for Legged Robots Using Vertex-Based ZMP Constraints, in IEEE Robotics and Automation Letters, 2(4), 2201-2208.
Hybrid direct collocation and control in the constraint-consistent subspace for dynamic legged robot locomotion
Pardo Diego, Neunert Michael, Winkler Alexander, Grandia Ruben, Buchli Jonas (2017), Hybrid direct collocation and control in the constraint-consistent subspace for dynamic legged robot locomotion, in Robotics: Science and Systems 2017.
Trajectory Optimization Through Contacts and Automatic Gait Discovery for Quadrupeds
Neunert Michael, Farshidian Farbod, Winkler Alexander W., Buchli Jonas (2017), Trajectory Optimization Through Contacts and Automatic Gait Discovery for Quadrupeds, in IEEE Robotics and Automation Letters, 2(3), 1502-1509.
Mobile robotic fabrication at 1:1 scale: the In situ FabricatorSystem, experiences and current developments
Giftthaler Markus, Sandy Timothy, Dörfler Kathrin, Brooks Ian, Buckingham Mark, Rey Gonzalo, Kohler Matthias, Gramazio Fabio, Buchli Jonas (2017), Mobile robotic fabrication at 1:1 scale: the In situ FabricatorSystem, experiences and current developments, in Construction Robotics, 000.
Learning motions from demonstrations and rewards with time-invariant dynamical systems based policies
Rey Joel, Kronander Klas, Farshidian Farbod, Buchli Jonas, Billard Aude (2017), Learning motions from demonstrations and rewards with time-invariant dynamical systems based policies, in Autonomous Robots, 000.
An efficient optimal planning and control framework for quadrupedal locomotion
Farshidian Farbod, Neunert Michael, Winkler Alexander W., Rey Gonzalo, Buchli Jonas (2017), An efficient optimal planning and control framework for quadrupedal locomotion, in 2017 IEEE International Conference on Robotics and Automation (ICRA), Singapore, Singapore.
Design, development and experimental assessment of a robotic end-effector for non-standard concrete applications
Kumar Nitish, Hack Norman, Doerfler Kathrin, Walzer Alexander Nikolas, Rey Gonzalo Javier, Gramazio Fabio, Daniel Kohler Matthias, Buchli Jonas (2017), Design, development and experimental assessment of a robotic end-effector for non-standard concrete applications, in 2017 IEEE International Conference on Robotics and Automation (ICRA), Singapore, Singapore.
Efficient kinematic planning for mobile manipulators with non-holonomic constraints using optimal control
Giftthaler Markus, Farshidian Farbod, Sandy Timothy, Stadelmann Lukas, Buchli Jonas (2017), Efficient kinematic planning for mobile manipulators with non-holonomic constraints using optimal control, in 2017 IEEE International Conference on Robotics and Automation (ICRA), Singapore, Singapore.
Online walking motion and foothold optimization for quadruped locomotion
Winkler Alexander W., Farshidian Farbod, Neunert Michael, Pardo Diego, Buchli Jonas (2017), Online walking motion and foothold optimization for quadruped locomotion, in 2017 IEEE International Conference on Robotics and Automation (ICRA), Singapore, Singapore.
Fast derivatives of rigid body dynamics for control, optimization and estimation
Neunert Michael, Giftthaler Markus, Frigerio Marco, Semini Claudio, Buchli Jonas (2016), Fast derivatives of rigid body dynamics for control, optimization and estimation, in 2016 IEEE International Conference on Simulation, Modeling, and Programming for Autonomous Robots (S, San Francisco, CA, USA.
Numerical search for local (partial) differential flatness
Sferrazza Carmelo, Pardo Diego, Buchli Jonas (2016), Numerical search for local (partial) differential flatness, in 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Daejeon, South Korea.
On reachability sets for optimal feedback controllers: Monitoring the approach of a region of attraction
Vomel Christof, Pardo Diego, Buchli Jonas (2016), On reachability sets for optimal feedback controllers: Monitoring the approach of a region of attraction, in 2016 IEEE International Conference on Robotics and Automation (ICRA), Stockholm.
A Projection Approach to Equality Constrained Sequential Linear Quadratic Optimal Control
Giftthaler Markus, Buchli Jonas, A Projection Approach to Equality Constrained Sequential Linear Quadratic Optimal Control, in 2017 IEEE-RAS International Conference on Humanoid Robots.
Dynamically Decoupling Base and End-Effector Motion for Mobile Manipulation using Visual-Inertial Sensing
Sandy Timothy, Buchli Jonas, Dynamically Decoupling Base and End-Effector Motion for Mobile Manipulation using Visual-Inertial Sensing, in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE 2017.
Real-Time Motion Planning of Legged Robots: A Model Predictive Control Approach
Farshidian Farbod, Jelavic Edo, Satapathy Asutosh, Giftthaler Markus, Buchli Jonas, Real-Time Motion Planning of Legged Robots: A Model Predictive Control Approach, in IEEE-RAS International Conference on Humanoid Robots (Humanoids).
Robust Whole-Body Motion Control of Legged Robots
Farshidian Farbod, Jelavic Edo, Winkler Alexander, Buchli Jonas, Robust Whole-Body Motion Control of Legged Robots, in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE 2017.

Associated projects

Number Title Start Funding scheme
160592 NCCR Robotics: Intelligent Robots for Improving the Quality of Life (phase II) 01.12.2014 National Centres of Competence in Research (NCCRs)
138920 The Swiss Robotic Disaster Recovery Challenge - Towards versatile robotic locomotion and manipulation 01.09.2012 SNSF Professorships
141853 NCCR Digital Fabrication: Advanced Building Processes in Architecture (phase I) 01.06.2014 National Centres of Competence in Research (NCCRs)

Abstract

We aim at formulating an architecture for simultaneously learning robust closed loop controllers and optimal plans with an emphasis on the application to the high dimensional nonlinear, hybrid and stochastic continous state-action spaces found in robots with arms and legs. We propose to use a combination of robust nonlinear control and reinforcement learning. As key contributions, in this project we will develop two main tools to solve this challenging control problem: (1) a robust iterative nonlinear optimal control method and (2) online certification methods that can be used to assess new candidate controllers before they are being rolled out on the real system. We will integrate these methods into a comprehensive learning robust control framework for a high performance torque controlled quadruped robot with arms. Herewith we address an important gap for the application of learning control to real demanding robotic control problems. While we gear our application and tests towards robots with arms and legs this research has large benefits for many other control problems in robotics (e.g. agile autonomous flight of quadcopter) and beyond.
-