Hierarchical motor learning and synthesis with passivity-based controller and phase oscillator
ABSTRACT In this paper, we propose a simple framework for learning and synthesis of fast and complex motor tasks. Where a passivity-based task-space controller acts not only as a full-body force control module, but also as an important module to generate phasic joint patterns. The generated joint patterns are encoded into the parameters of phase oscillators and form the synergy of the task. Then, similar and/or faster motions are synthesized by superposing the task space controller output and the oscillator output with the modified oscillator amplitudes and/or frequencies. We present some examples of whole-body motion synthesis on a human-sized biped humanoid robot including squatting, dancing and stepping while bipedal balancing. The simulation and experimental videos are supplemented.
- SourceAvailable from: Yohei Otaka
- "Typical behavior of superposition of SB and DB (Video 7) Fig. 12. Combined push-recovery with stepping (Video 7) used. However, dynamic motions are expected to be learned by robot itself . Our future effort will be devoted to this work. "
Conference Paper: Integration of multi-level postural balancing on humanoid robots[Show abstract] [Hide abstract]
ABSTRACT: This paper discusses an integration issue of multi-level postural balancing on humanoid robot. We give a unified viewpoint of postural balancing, which covers Ankle Strategy to Hip Strategy. Two kinds of distributor of desired ground reaction force to whole-body joint torque are presented. The one distributor leads to a dynamic balancer which covers Hip strategy, with the under-actuated situation. A simple angular momentum regulator is also proposed to stabilize the internal motions due to the joint redundancy. The other distributor leads to a static balancer which lies between Ankle and Hip strategy. Furthermore, this paper demonstrates that replacement of the center of mass feedback with the local joint stiffness makes the robot much stabler for some fast motions. Motivated by the practicability of the static balancer and the strong push-recovery performance of the dynamic balancer, this paper presents a simple integration by superposition of the both balancers on a compliant human-sized biped robot. The simulation and experimental videos are supplemented.Robotics and Automation, 2009. ICRA '09. IEEE International Conference on; 06/2009
- [Show abstract] [Hide abstract]
ABSTRACT: We propose a new supervised learning and syn- thesis framework for fast and complex motor tasks. Wherein, a statics-based task-space controller acts not only as a full- body motion control module, but also as a module to generate synergetic joint patterns. The generated joint patterns are encoded into the parameters of phase trajectories of attractors and form the synergy of the task. Similar, but faster motions are synthesized by superposing the task-space controller output and the trajectory attractor output with the modified parameters, while learning dynamics and stiffness according to the task error. We demonstrate the proposed framework by simulating a balanced fast squat on a humanoid robot model.
Conference Paper: Humanoid batting with bipedal balancing[Show abstract] [Hide abstract]
ABSTRACT: This paper reports our first attempt to achieve a baseball batting demonstration with a human-sized bipedal humanoid robot, aimed at presenting a performance integrating perception, control and learning. Real-time whole-body motion control and visual perception are integrated to allow the robot to predict the ball position and hit it. The ball was thrown by a human and recognized by the eye cameras. For the prediction, we propose a simple sequential estimator to predict the arrival time and position of the ball. For the control, fast and smooth batting trajectories are superposed on a whole-body force controller taking account of bipedal balancing. Although the prediction and learning model are not fully implemented, this paper demonstrates promising simulation and experimental results with the proposed framework. So far, we have succeeded at the timing prediction and fast swing motion within 300 ms without falling. Experimental and simulation videos are available as supplementary material.Humanoid Robots, 2008. Humanoids 2008. 8th IEEE-RAS International Conference on; 01/2009