INTELLIGENT CONTROL OF DaNI ROBOT BASED ON ROBOT VISION AND OBJECT RECOGNITION

In order to use the mobile robot platform for tracking particular moving objects it is necessary to develop intelligent top level control algorithm. In this paper a computational intelligence based object recognition algorithm in robot vision system is presented. The main goal of this research was to enable DaNI robot to recognize the particular Lego NXT robot among other differently shaped Lego NXT robots and objects, and to localize them with accuracy high enough to allow following the chosen one. The necessary robustness of the 2D object recognition is achieved by a novel robust robot vision systems that introduces the closed-loop control of image segmentation without the use of extensive previous knowledge and computational intelligence as an important part of the vision system. Reliable feature extraction is necessary to fully exploit intelligent classifiers, which are the core of the proposed 2D object recognition method. Two different types of classifiers were developed and then compared, the ANFIS neuro-fuzzy classifier and neural network classifier. The supervisory control algorithm was tested through experiments and the results are showing that this kind of approach in robot vision for object recognition provides good results. Besides that, future use of computational intelligence techniques in robotic vision object recognition system is also briefly discussed.