Workshop: Learning for Locomotion
Quick Facts
| Organizers: | Jan Peters, Russ Tedrake, Stefan Schaal |
| Conference: | Robotics - Science and Systems 2005 |
| Date: | June 11, 2005 |
| Room: | To be announced |
| Location: | MIT, Cambridge, MA, USA |
| Website: | http://www.jan-peters.net/Research/LearningForLocomotion |
Abstract
Over the last few decades, there has been an impressive amount of published work on legged locomotion, including bipedal walking, running, hopping, stand-ups, summersaults and much more. However, despite all this progress, legged locomotion research has largely been driven by researchers using human insight and creativity in order to generate locomotion control algorithms. In order to improve the robustness, energy efficiency, and natural appearance of legged locomotion, there may be a significant advantage to using machine learning methods to synthesize new controllers and to avoid tedious parameter tuning. For instance, it could be advantageous to learn dynamics models, kinematic models, impact models, for model-based control techniques. Imitation learning could be employed for the teaching of gaits patterns, and reinforcement learning could help tuning parameters of the control policies in order to improve the performance with respect to given cost functions. In this context, we would like to bring together researchers from both the legged locomotion and machine learning in order to discuss which locomotion problems require learning, and to identify the machine learning methods that can be used to solve them.
Goal
In order to better understand the application of machine learning techniques to locomotion, our goal is to bring together researchers who represent many different approaches to biped locomotion control with their peers in machine learning for control. We hope to discuss future research directions for principled machine learning approaches to biped locomotion. The workshop will address topics such as:
- Which unsolved biped locomotion problems can be solved using learning?
- Can walking be broken down into components upon which machine learning methods are applicable?
- What models (e.g., forward, inverse, impact) would be desirable for controlling locomotion?
- Can machine learning methods help solve the gait generation and foot-placement problems?
- Can human learning of locomotion yield insights for both robotics and machine learning?
- Which machine learning algorithms are suitable for online implementation on the robot, and which problems can be solved in simulation?
- What cost functions should be used to describe "optimal" walking, and what experiments should be done to test our controllers?
Furthermore, we intend to kick-off the Legged Robot Control Competition.
Format
The workshop is 3.5 hours in the morning, and 2.5 hours in the afternoon. The invited participants will give short (20-25 minute) talks on particular aspects of real-time modeling. After each talk there will be a fair amount of discussion time and additionally the possibility for other participants to give short talks (less then 10 min) that complement the topic of the main talks. Participants may be allowed to give multiple short talks and the possibility of impromptu talks from members of the audience could also be considered.
Tentative Program
| 8:25-8:30am | Welcome |
| Jan Peters (USC), Russ Tedrake (MIT), Stefan Schaal (USC) |
Session 1: Control for Legged Robots
| 8:30-8:55am | Adaptive control of locomotion in articulated robots using central pattern generators |
| Auke Ijspeert, EPFL | |
| 8:55-9:20am | Control of Dynamic Legged Robots |
| Marc Raibert and Martin Buehler, Boston Dynamics | |
| 9:20-9:45am | The Design and Control of Bio-inspired Legged Robots |
| Chandana Paul, Cornell University | |
| 9:45-10:10am | Biomechanics Teaches Us About Biped Control Design |
| Hugh Herr, Massachuesetts Institute of Technology | |
| 10:10-10:35am | Nonlinear stability tools for locomotion and learning |
| Jean-Jacques Slotine, Massachuesetts Institute of Technology |
Coffee Break (10:35-10:45am)
Session 2: Funding Opportunities and Competitions
| 10:45-11:10am | DARPA's Learning Locomotion Program |
| Eric Krotkov on behalf of Larry Jackel, DARPA Information Processing Technology Office | |
| 11:10-11:15am | Legged Robot Control Competition: Kick-Off |
| Russ Tedrake, Massachusetts Institute of Technology |
Short Break (11:15-11:20am)
Session 3: Systematic and Automatic Gait Generation
| 11:20-11:45pm | Systematic creation of stable walking and running gaits in planar bipeds |
| Eric Westervelt, Ohio State University | |
| 11:45-12:10pm | Locomotion on the Vertical: Early Efforts in Gait Development for Robot Climbing |
| Alfred Rizzi, Carnegie Mellon University | |
| 12:10-12:35pm | Learning and Adaptation of Biped Locomotion with Dynamical Movement Primitives |
| Jun Nakanishi, Advanced Telecommunication Research (ATR) Institute |
Lunch Break (12:35-2:05pm)
Session 4: Planning and Optimization for Locomotion
| 2:05-2:25pm | Optimizing in Low-Dimensional, Behavior-Specific Spaces |
| Jessica Hodgkins, Carnegie Mellon University | |
| 2:25-2:45pm | Co-evolutionary learning for robotic locomotion |
| Hod Lipson, Cornell University |
Session 5: Locomotion from Reinforcement Learning
| 2:45-3:10pm | Quadruped locomotion via reinforcement learning |
| Andrew Ng, Stanford University |
Coffee Break (3:10-3:20pm)
Session 5 continued: Locomotion from Reinforcement Learning
| 3:20-3:45pm | Robust reasoning for robot planning |
| Geoff Gordon, Carnegie Mellon University | |
| 3:45-4:10pm | Model-based Reinforcement Learning |
| Christopher G. Atkeson, Carnegie Mellon University | |
| 4:10-4:35pm | Model-based and Model-free reinforcement learning methods for biped walking |
| Jun Morimoto, Advanced Telecommunication Research (ATR) Institute | |
| 4:35-5:00pm | Exploiting passive dynamics to achieve fast online policy learning |
| Russ Tedrake, Massachuesetts Institute of Technology |
---
2. Tentative Program
| 8:45-8:50am | Welcome |
| Jan Peters (USC), Russ Tedrake (MIT), Stefan Schaal (USC) |
Session 1: Systematic and Automatic Gait Generation
| 8:50-9:10am | Systematic creation of stable walking and running gaits in planar bipeds |
| Eric Westervelt, Ohio State University | |
| 9:10-9:30am | Locomotion on the Vertical: Early Efforts in Gait Development for Robot Climbing |
| Alfred Rizzi, Carnegie Mellon University | |
| 9:30-9:50am | Learning and Adaptation of Biped Locomotion with Dynamical Movement Primitives |
| Jun Nakanishi, Advanced Telecommunication Research (ATR) Institute |
Short Break (9:50-9:55am)
Session 2: Planning and Optimization for Locomotion
| 9:55-10:15am | Optimizing in Low-Dimensional, Behavior-Specific Spaces |
| Jessica Hodgkins, Carnegie Mellon University | |
| 10:15am-10:35pm | Co-evolutionary learning for robotic locomotion |
| Hod Lipson, Cornell University |
Coffee Break (10:35-10:45am)
Session 3: Funding Opportunities and Competitions
| 10:45-11:05am | DARPA's Learning Locomotion Program |
| Eric Krotkov on behalf of Larry Jackel, DARPA Information Processing Technology Office | |
| 11:05-11:10am | Legged Robot Control Competition: Kick-Off |
| Russ Tedrake, Massachusetts Institute of Technology |
Short Break (11:10-11:15am)
Session 4a: Control for Legged Robots - Part I
| 11:15-11:40am | Control of Dynamic Legged Robots |
| Marc Raibert and Martin Buehler, Boston Dynamics | |
| 11:40-12:00pm | Nonlinear stability tools for locomotion and learning |
| Jean-Jacques Slotine, Massachuesetts Institute of Technology |
Lunch Break (12:00-1:30pm)
Session 4b: Control for Legged Robots - Part II
| 1:30-1:50pm | Biomechanics Teaches Us About Biped Control Design |
| Hugh Herr, Massachuesetts Institute of Technology | |
| 1:50-2:10pm | The Design and Control of Bio-inspired Legged Robots |
| Chandana Paul, Cornell University | |
| 2:10-2:30pm | Adaptive control of locomotion in articulated robots using central pattern generators |
| Auke Ijspeert, EPFL |
Short Break (2:30-2:35pm)
Session 5a: Locomotion from Reinforcement Learning
| 2:35-2:55pm | Quadruped locomotion via reinforcement learning |
| Andrew Ng, Stanford University | |
| 2:55-3:15pm | Model-based and Model-free reinforcement learning methods for biped walking |
| Jun Morimoto, Advanced Telecommunication Research (ATR) Institute |
Coffee Break (3:15-3:25pm)
Session 5b: Locomotion from Reinforcement Learning
| 3:25-3:45pm | Robust reasoning for robot planning |
| Geoff Gordon, Carnegie Mellon University | |
| 3:45-4:05pm | Model-based Reinforcement Learning |
| Christopher G. Atkeson, Carnegie Mellon University | |
| 4:05-4:25pm | Exploiting passive dynamics to achieve fast online policy learning |
| Russ Tedrake, Massachuesetts Institute of Technology |
Participants
This workshop will bring together researchers from both the robotics and machine learning in order to explore how to approach the topic of learning legged locomotion in a principled way. Participants of the workshop (inclusive of the audience) are encouraged to actively participate by responding with questions and comments about the talks and give stand-up talks. Please contact the organizers if you would like to reserve apriori some time for expressing your view (short short-talk!!) on a particular topic.
Organizers
The workshop is organized by Jan Peters, Russ Tedrake and Stefan Schaal, from the Departments of Computer Science and Neuroscience, University of Southern California, Los Angeles, CA, USA and from the Brain and Cognitive Sciences Department at the Massachusetts Institute of Technology, Cambridge, MA, USA.
Location and More Information
The most up-to-date information about Robotics - Science and Systems 2005 can be found on the Robotics 2005 website.