Model based control approaches for multi-modal locomotion of a humanoid robot
Event details
| Date | 05.12.2014 |
| Hour | 14:00 › 15:00 |
| Speaker | Salman Faraji |
| Location | |
| Category | Conferences - Seminars |
Performing locomotion is a very normal task for human in every-day life, though very difficult for robots. Although there are many powerful humanoids built and equipped with different kinds of sensors and actuators, performing complex and agile locomotion like humans is far beyond existing algorithms and methods. Such performance requires powerful actuators, fast perception, whole body planning and intrinsic compliance of the hardware and the control method in order to minimize damage risks and ensure robustness and efficiency. In this presentation, we briefly review the literature regarding locomotion of humanoid robots and animated characters. We break down the control loop into several blocks and mention the state of the art, concerning the real hardware available in BioRob laboratory. We focus then on control blocks in different layers and propose a hierarchical architecture that decouples instantaneous control and future planning, using a simplified model called Inverted Pendulum. Such a fast planning method enables the robot to perform walking at reasonable speeds but with certain limitations, mainly producing less human-like motion. This approach intensively uses quadratic programming for state estimation, inverse dynamics and planning which are the main blocks in the proposed control architecture. We describe details of each from theoretical perspective and wrap up the presentation by discussing possible improvements and extensions for this architecture to include the full body for more complex locomotion types.
CV
Salman Faraji received his bachelor degree in Electrical Engineering from Sharif University in Iran. He further performed a master of Manufacturing systems and Robotics in EPFL university, Switzerland. His master thesis was about performing bipedal walking on a simulated torque-controlled robot, supervised by Christopher Atkeson in Carnegie Mellon university and Auke Jan Ijspeert in EPFL who supervises him as a doctoral assistant now in BioRob laboratory. He is contributing to the European Walk-Man project and developing whole-Body agile locomotion algorithms.
CV
Salman Faraji received his bachelor degree in Electrical Engineering from Sharif University in Iran. He further performed a master of Manufacturing systems and Robotics in EPFL university, Switzerland. His master thesis was about performing bipedal walking on a simulated torque-controlled robot, supervised by Christopher Atkeson in Carnegie Mellon university and Auke Jan Ijspeert in EPFL who supervises him as a doctoral assistant now in BioRob laboratory. He is contributing to the European Walk-Man project and developing whole-Body agile locomotion algorithms.
Practical information
- Expert
- Free