Towards a Line-Crawling Robot Obstacle Classification System: A Rough Set Approach

The basic contribution of this paper is the presentation of two methods that can be used to design a practical robot obstacle classification system based on data mining methods from rough set theory. These methods incorporate recent advances in rough set theory related to coping with the uncertainty in making obstacle classification decisions either during the operation of a mobile robot. Obstacle classification is based on the evaluation of data acquired by proximity sensors connected to a line-crawling robot useful in inspecting power transmission lines. A fairly large proximity sensor data set has been used as means of benchmarking the proposed classification methods, and also to facilitate comparison with other published studies of the same data set. Using 10-fold cross validated paired t-test, this paper compares the rough set classification learning method with the Waikato Environment for Knowledge Analysis (WEKA) classification learning method.