Workshop on Numerical optimization for Online multi-contact Motion Planning and control

Robotics: Science and Systems - June 22, 2019 - Freiburg, Germany

Location:

Faculty of Engineering, Georges-Köhler-Allee 101, 79110 Freiburg im Breisgau

Building 101, Room 01 009

What if legged robots were able to generate dynamic motions in real-time while interacting with a complex environment? Such technology would represent a significant step forward the deployment of legged systems in real world scenarios. This means being able to replace humans in the execution of dangerous tasks and to collaborate with them in industrial applications.

Numerical optimization and data-driven algorithms can help us to tackle this challenge and enable motion planning and control for legged robotic systems in complex geometry environments (e.g. multi-contact scenarios). Indeed, when the complexity of the terrain increases, or when the execution of the requested task involves highly dynamic motions, numerical optimization and machine learning strategies are needed to automatically find feasible trajectories and control actions that could not otherwise be determined.

The presence of obstacles, possible disturbances and/or modeling errors makes it necessary to find those control policies in the order of milliseconds so that the robot can immediately compensate unexpected events and re-plan suitable reactions.

This workshop aims to bring together researchers from all the relevant communities in legged locomotion such as: numerical optimization, machine learning (ML), model predictive control (MPC) and computational geometry in order to chart the most promising methods to address the above-mentioned scientific challenges.

• Numerical optimization for legged locomotion
• Fast (real-time or online) non-linear MPC and optimal control
• Trajectory optimization through contacts
• Computational geometry for generation loco-manipulation feasible motions
• Multi-contact motion planning and control for highly dynamic and dexterous maneuvers
• Accurate and soft contact modeling for numerical optimization
• Combination of optimal control and machine learning formalism for data exploration and exploitation
• Synchronization and blending of fast real-time control policies with slow online motion planning

Speaker Videos Slides Title

9.00 - 9.10 Organizers Introductory remarks

9.10 - 9.40 Emo Todorov 0h0m01s slides Acceleration-based methods for trajectory optimization through contacts

9.40 - 10.05 Daniele Pucci 0h25m30s slides Beyond torque control: towards jerk control of humanoid robots via contact parametrisation

10.05 - 10.30 Kris Hauser 0h52m30s Real time optimization for humanoid fall protection

10.30 - 11.00 coffee break

11.00 - 11.25 Denis Teissander 1h45m00s Convex polyhedra and the world of screws, tolerancing analysis and kinematics

11.25 - 11.50 Vincent Delos 2h10m40s Efficient computations of sums and intersections of high dimensional convex polytopes

11.50 - 12.10 Romeo Orsolino 2h30m40s slides Actuation-aware simplified dynamic models for robotic legged locomotion

12.15 - 13.00 Lunch break

13.00 - 13.30 Moritz Diehl 3h46m20s slides Recent advances on numerical optimal control with state dependent switches

13.30 - 13.50 Andrea Zanelli 4h16m00s Efficient Inexact methods for NMPC with Guarantees and their high-performance implementation in acados

13.50 - 14.20 Andrea Del Prete 4h34m50s slides Multi-contact balancing on elastic contacts

14.20 - 14.50 Justin Carpentier 5h5m50s Analytical derivatives of contact dynamic equations and their application to legged locomotion

14.50 - 15.10 Poster spotlights

15.15 - 16.00 Coffee break + poster session (in the Robotics Hall)

16.00 - 16.25 Carlos Mastalli 6h53m00s slides Highly dynamic maneuvers computed by feasible-prone DDP

16.25 - 16.55 Ludovic Righetti 7h16m00s slides Efficient computation of multi-contact movements

16.55 - 17.30 Nicolas Mansard 7h47m00s Final panel discussion (session chaired by Dr. Nicolas Mansard)

The workshop is soliciting posters/videos for an interactive session and a 2-min teaser talk for all participants. To participate, please submit a poster title and an 2 pages abstract (using IEEE two-column format) via email with the subject "[RSS 2019 Abstract]" to: (rss2019numoptmulticontact@gmail.com) by May 29, 2019 (AoE). All contributions will undergo a brief review by the organizers, and the authors will be notified of acceptance by June 9, 2019. Please note that we have a best poster award.

Accepted papers and eventual supplementary material will be made available on the workshop website. However, this does not constitute an archival publication and no formal workshop proceedings will be made available, meaning contributors are free to publish their work in archival journals or conference.

This workshop is supported by MEMMO project. MEMMO is a collaborative project supported by European Union within the H2020 Program, under Grant Agreement No. 780684. The organizers would like to thank also the Dr. Semini for his valuable feedback.

At the following link you will find a Google Doc where you can write possible questions/topics that, in your opinion, should be discussed during the final speakers' panel:

• https://docs.google.com/spreadsheets/d/17N1uIiMCBk9vKN1HDZuUE9um77f6mRAousr-j8htRlQ/edit?usp=sharing

The workshop is going to be streamed live through youtube. You can watch the event live at this link .