Instance-based reinforcement learning technique with a meta-learning mechanism for robust multi-robot systems

Authors:
Toshiyuki Yasuda;Motohiro Wada;Kazuhiro Ohkura
Affiliations:
Hiroshima University, Higashi-Hiroshima, Japan;Hiroshima University, Higashi-Hiroshima, Japan;Hiroshima University, Higashi-Hiroshima, Japan
Venue:
TAROS'11 Proceedings of the 12th Annual conference on Towards autonomous robotic systems
Year:
2011

Citing 11
Cited 0

Introduction to Reinforcement Learning

Introduction to Reinforcement Learning
Metalearning and neuromodulation

Neural Networks - Computational models of neuromodulation
Acquisition of Stand-up Behavior by a Real Robot using Hierarchical Reinforcement Learning

ICML '00 Proceedings of the Seventeenth International Conference on Machine Learning
Meta-learning in reinforcement learning

Neural Networks
Reinforcement Learning in Continuous Time and Space

Neural Computation
Natural Actor-Critic

Neurocomputing
A Reinforcement Learning Technique with an Adaptive Action Generator for a Multi-robot System

SAB '08 Proceedings of the 10th international conference on Simulation of Adaptive Behavior: From Animals to Animats
Co-evolution of Shaping Rewards and Meta-Parameters in Reinforcement Learning

Adaptive Behavior - Animals, Animats, Software Agents, Robots, Adaptive Systems
Regularization and feature selection in least-squares temporal difference learning

ICML '09 Proceedings of the 26th Annual International Conference on Machine Learning
Improving search efficiency in the action space of an instance-based reinforcement learning technique for multi-robot systems

ECAL'07 Proceedings of the 9th European conference on Advances in artificial life
Dimension reduction and its application to model-based exploration in continuous spaces

Machine Learning

Quantified Score

Hi-index	0.00

Visualization

Abstract

In recent years, the subject of learning autonomous robots has been widely discussed. Reinforcement learning (RL) is a popular method in this domain. However, its performance is quite sensitive to the discretization of state and action spaces. To overcome this problem, we have developed a new technique called Bayesian-discriminationfunction-based RL (BRL). BRL has proven to be more effective than other standard RL algorithms in dealing with multi-robot system (MRS) problems. However, similar to most learning systems, BRL occasionally suffers from overfitting. This paper introduces an extension of BRL for improving the robustness of MRSs. Meta-learning based on the information entropy of firing rules is adopted for adaptively modifying its learning parameters. Physical experiments are conducted to verify the effectiveness of our proposed method.