Learning Robot Gait Stability using Neural Networks as Sensory Feedback Function for Central Pattern Generators

Sébastien Gay, Auke Ijspeert, José Santos-Victor
2013
Article de conférence

In this paper we present a framework to learn a model-free feedback controller for locomotion and balance control of a compliant quadruped robot walking on rough terrain. Having designed an open-loop gait encoded in a Central Pattern Generator (CPG), we use a neural network to repre- sent sensory feedback inside the CPG dynamics. This neural network accepts sensory inputs from a gyroscope or a camera, and its weights are learned using Particle Swarm Optimization (unsupervised learning). We show with a simulated compliant quadruped robot that our controller can perform significantly better than the open-loop one on slopes and randomized height maps.

À propos de ce résultat

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Concepts associés

Chargement

Publications associées

Chargement

Sébastien Gay, Auke Ijspeert, José Santos-Victor
2013
Article de conférence

Résumé

Official source

https://infoscience.epfl.ch/record/187784?ln=fr

À propos de ce résultat

Concepts associés (14)

Concepts associés

Chargement

Publications associées (19)

Chargement

Afficher plus

Publications associées

Chargement

Réseau locomoteur spinal

Le réseau locomoteur spinal, ou central pattern generator (CPG) en anglais, est un réseau de neurones localisé dans la moelle spinale responsable de la locomotion. La particularité de ce réseau est

Neural network

A neural network can refer to a neural circuit of biological neurons (sometimes also called a biological neural network), a network of artificial neurons or nodes in the case of an artificial neur

Robotique

thumb|upright=1.5|Nao, un robot humanoïde. thumb|upright=1.5|Des robots industriels au travail dans une usine. La robotique est l'ensemble des techniques permettant la conception et la réalisation de

Learning Robot Gait Stability using Neural Networks as Sensory Feedback Function for Central Pattern Generators

Sébastien Gay, Auke Ijspeert, José Santos-Victor

In this paper we present a framework to learn a model-free feedback controller for locomotion and balance control of a compliant quadruped robot walking on rough terrain. Having designed an open-loop gait encoded in a Central Pattern Generator (CPG), we use a neural network to represent sensory feedback inside the CPG dynamics. This neural network accepts sensory inputs from a gyroscope or a camera, and its weights are learned using Particle Swarm Optimization (unsupervised learning). We show with a simulated compliant quadruped robot that our controller can perform significantly better than the open-loop one on slopes and randomized height maps.

Ieee2013

Chargement

Learning to move in modular robots using central pattern generators and online optimization

Auke Ijspeert, Jérôme Maye

This article addresses the problem of how modular robotics systems, i.e. systems composed of multiple modules that can be configured into different robotic structures, can learn to locomote. In particular, we tackle the problems of online learning, that is, learning while moving, and the problem of dealing with unknown arbitrary robotic structures. We propose a framework for learning locomotion controllers based on two components: a central pattern generator (CPG) and a gradient-free optimization algorithm: Powell’s method. The CPG is implemented as a system of coupled nonlinear oscillators in our YaMoR modular robotic system, with one oscillator per module. The nonlinear oscillators are coupled together across modules using Bluetooth communication to obtain specific gaits, i.e. synchronized patterns of oscillations among modules. Online learning is done by running the Powell optimization algorithm in parallel to the CPG model, with the speed of locomotion being the criterion to be optimized. Interesting aspects of the optimization are: it is carried out online, it does not require stopping or resetting the robots, and it is fast. We present results showing the interesting properties of this framework for a modular robotic system. In particular, our CPG model can readily be implemented in a distributed system, it is cheap computationally, it exhibits limit cycle behavior (temporary perturbations are rapidly forgotten), it produces smooth trajectories even when control parameters are abruptly changed, and it is robust against imperfect communication among modules. We also present results of learning to move with three different robot structures. Interesting locomotion modes are obtained after running the optimization for less than 60 minutes.

2008

Chargement

Trot Gait Locomotion of A Cat Sized Quadruped Robot

Mostafa Ajallooeian, Auke Ijspeert, Alexander Spröwitz, Alexandre Tuleu

We are proposing and testing two model-free approaches for locomotion control of a light-weight, compliant, quadruped robot: open loop central pattern generators (CPG), and open and closed- loop dynamical movement primitives (DMP). We are presenting two different knee joint controllers, based on the hypothesis that the passive- compliant leg design might require less control effort for the knee joint control. CPG-control parameters are optimized applying extensive optimization runs using PSO, resulting trot gaits are evaluated for speed and robustness. One derived gait is selected as learning input for a gait controller based on DMP. We design the DMP-based trot gait controller both for open and closed-loop control. For the latter we are feeding back gyroscope and touch sensor signals to modulate the core phase pattern. Found CPG based trot gaits are up to 90cm/sec fast (more than 3 BL/sec). The joint controller assisting the passive compliant knee joint shows gaits with higher speed, and larger hip amplitude. Closed- loop DMP based gaits let the robot handle frontal and sideways slopes, step-downs, and random pushes to the robot’s COM more robustly.

2011

Chargement

Afficher plus