Paper Info
Title
First Results from Further Experimental Comparisons between Pittsburgh and Michigan Fuzzy Classifier Systems for Mobile Robotics
Abstract
This paper presents new work carried out in simulation on a performance comparison between the Michigan and Pittsburgh Fuzzy Classifier System approaches to a control problem in mobile robotics. First results from this work, combined with thoughts on our previous work, indicate that changes in the problem formulation can swing the balance between whether the Michigan or Pittsburgh approaches give the best results, in terms of fast and robust convergence on high performance solutions. In the context of the evolutionary and learning tasks overall, the changes could appear minor, in this case modifying the sensor range of the robot. However, such changes can fundamentally modify the processes of a fuzzy controller in a given environment and, thereby, the characteristics of fitness evaluations. However, these early results are not yet statistically significant, a larger experimental programme is required to achieve this.
Year
Venue
Keywords
2005
EUSFLAT Conf.
fuzzy logic,genetic algorithms,mobile robotics,learning classifier systems,mobile robot,learning classifier system,genetic algorithm,statistical significance
Field
DocType
Citations 
Convergence (routing),Control theory,Fuzzy set operations,Computer science,Fuzzy logic,Artificial intelligence,Robot,Fuzzy classifier,Machine learning,Robotics,Swing
Conference
0
PageRank 
References 
Authors
0.34
6
2
Name
Order
Citations
PageRank
Anthony G. Pipe125539.08
Brian Carse225926.31