Publication: Performance comparison between a generic and specialized artificial neural system for robotic fault recognition
Loading...
Date
2009-04-01
Authors
Teo, Yu Xiang
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Nowadays, the autonomous robots are widely used in many kinds of industrial field. The latest technology enables the design of advanced and complex robot. As a result of the increasing of complexity, the probability of the occurring of system faults will increase proportional too. If the faults do not have any proper solutions, these will lead to the calamitous consequences especially for the autonomous robots. Hence, there is a growing need for the system for detection, recognition, classification and find solution in order to handle the fault as soon as it occurs. Artificial Neural Network (ANN) can be used for the detection and classification of the robot faults. In this project, generic and specialized neural systems are constructed using the competitive learning in order to do the classification of two types of faults, i.e. abnormal robot movement and normal (no faults). Performances of both generic and specialized neural systems are compared. The percentage of correctness for generic neural system is Normal class, 73.85%; Collision class, 69.92%; and Obstruction class, 29.09%. For specialized neural system, the percentage of correctness is 92.31%, 94.74%, and 96.36% for Normal, Collision and Obstruction classes, respectively. As a conclusion, specialized neural system has the better percentage of correctness than the generic neural system in the robot fault recognition.