Mechanism of failure diagnosis based on language games and q-learning over a termites simulator

Authors:
Arles Ernesto Rodriguez Portela;Jonatan Gómez Perdomo
Affiliations:
National University of Colombia, Bogota, Colombia;National University of Colombia, Bogota, Colombia
Venue:
Proceedings of the 13th annual conference companion on Genetic and evolutionary computation
Year:
2011

Citing 3
Cited 0

The Vision of Autonomic Computing

Computer
Collaborative Self-Configuration and Learning in Autonomic Computing Systems: Applications to Supply Chain

ICAC '06 Proceedings of the 2006 IEEE International Conference on Autonomic Computing
Ant colony system: a cooperative learning approach to the traveling salesman problem

IEEE Transactions on Evolutionary Computation

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a mechanism of failure diagnosis in a multi-agent environment of termites looking for food. The failure system is defined based on the probability that each termite (agent) has of executing a movement instruction. By using language games concepts and a version of the Q-learning algorithm, termites diagnose failures with the highest failure probabilities. Termites also have enough information to determine which movement actuator is failing based on a simple voting system that is the result of language games of diagnosis. Results show that the proposed approach is able, from local interactions, to build a set of very specific diagnosis questions, allowing the system to diagnose more than one type of failure at the same time, while the accounted number of diagnosis questions for instructions with low failure probability is reduced.