Acquisition of a Biped Walking Policy Using an Approximated Poincaré Map
نویسندگان
چکیده
We propose a model-based reinforcement learning algorithm for biped walking in which the robot learns to appropriately place the swing leg. This decision is based on a learned model of the Poincaré map of the periodic walking pattern. The model maps from a state at a single support phase and foot placement to a state at the next single support phase. We applied this approach to both a simulated robot model and an actual biped robot. Successful walking patterns are acquired.
منابع مشابه
Nonparametric representation of an approximated Poincaré map for learning biped locomotion
We propose approximating a Poincaré map of biped walking dynamics using Gaussian processes. We locally optimize parameters of a given biped walking controller based on the approximated Poincaré map. By using Gaussian processes, we can estimate a probability distribution of a target nonlinear function with a given covariance. Thus, an optimization method can take the uncertainty of approximated ...
متن کاملStable Gait Planning and Robustness Analysis of a Biped Robot with One Degree of Underactuation
In this paper, stability analysis of walking gaits and robustness analysis are developed for a five-link and four-actuator biped robot. Stability conditions are derived by studying unactuated dynamics and using the Poincaré map associated with periodic walking gaits. A stable gait is designed by an optimization process satisfying physical constraints and stability conditions. Also, considering...
متن کاملAnalysis of 3D Passive Walking Including Turning Motions for the Finite-width Rimless Wheel
The focus of studies in the field of passive walking has often been on straight walking, while less attention has been paid to the field of turning motions. In this paper, the passive motions of a finite width rimless wheel as the simplest 3D model of passive biped walkers was investigated with a focus on turning motions. For this purpose, the hybrid model of the system consisting of continuous...
متن کاملPoincaré-Map-Based Reinforcement Learning For Biped Walking
We propose a model-based reinforcement learning algorithm for biped walking in which the robot learns to appropriately modulate an observed walking pattern. Viapoints are detected from the observed walking trajectories using the minimum jerk criterion. The learning algorithm modulates the via-points as control actions to improve walking trajectories. This decision is based on a learned model of...
متن کاملRobust Trajectory Free Model Predictive Control of Biped Robots with Adaptive Gait Length
This paper employs nonlinear disturbance observer (NDO) for robust trajectory-free Nonlinear Model Predictive Control (NMPC) of biped robots. The NDO is used to reject the additive disturbances caused by parameter uncertainties, unmodeled dynamics, joints friction, and external slow-varying forces acting on the biped robots. In contrary to the slow-varying disturbances, handling sudden pushing ...
متن کامل