Item


Towards Direct Policy Search Reinforcement Learning for Robot Control

This paper proposes a high-level reinforcement learning (RL) control system for solving the action selection problem of an autonomous robot. Although the dominant approach, when using RL, has been to apply value function based algorithms, the system here detailed is characterized by the use of direct policy search methods. Rather than approximating a value function, these methodologies approximate a policy using an independent function approximator with its own parameters, trying to maximize the future expected reward. The policy based algorithm presented in this paper is used for learning the internal state/action mapping of a behavior. In this preliminary work, we demonstrate its feasibility with simulated experiments using the underwater robot GARBI in a target reaching task

© IEEE/RSJ International Conference on Intelligent Robots and Systems, 2006, p. 3178-3183

IEEE

Author: El-Fakdi Sencianes, Andrés
Carreras Pérez, Marc
Ridao Rodríguez, Pere
Date: 2006
Abstract: This paper proposes a high-level reinforcement learning (RL) control system for solving the action selection problem of an autonomous robot. Although the dominant approach, when using RL, has been to apply value function based algorithms, the system here detailed is characterized by the use of direct policy search methods. Rather than approximating a value function, these methodologies approximate a policy using an independent function approximator with its own parameters, trying to maximize the future expected reward. The policy based algorithm presented in this paper is used for learning the internal state/action mapping of a behavior. In this preliminary work, we demonstrate its feasibility with simulated experiments using the underwater robot GARBI in a target reaching task
Format: application/pdf
Citation: El-Fakdi, A., Carreras, M., i Ridao, P. (2006). Towards Direct Policy Search Reinforcement Learning for Robot Control. IEEE/RSJ International Conference on Intelligent Robots and Systems, 3178 - 3183. Recuperat 06 maig 2010, a http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4058885
ISBN: 1-4244-0258-1
Document access: http://hdl.handle.net/10256/2221
Language: eng
Publisher: IEEE
Collection: Reproducció digital del document publicat a: http://dx.doi.org/10.1109/IROS.2006.282342
Articles publicats (D-ATC)
Is part of: © IEEE/RSJ International Conference on Intelligent Robots and Systems, 2006, p. 3178-3183
Rights: Tots els drets reservats
Subject: Aprenentatge per reforç
Robots autònoms -- Sistemes de control
Autonomous robots -- Control systems
Reinforcement learning
Title: Towards Direct Policy Search Reinforcement Learning for Robot Control
Type: info:eu-repo/semantics/article
Repository: DUGiDocs

Subjects

Authors