Abstract
The visual navigation of mobile robots rely mostly on road edge detection from images taken from onboard camera. The algorithms used to detect road edges give globally satisfactory results if the scene variability is limited. However, sometimes even if the road edges are properly detected from the input image, the robot guidance task still fails. In this context, this paper investigates the subjective interaction between the scene model and the world model, and we propose a visual control scheme for robot guidance that minimizes the model error induced by processing raw image data. The involved control system includes the fuzzy approach at two levels: a fuzzy perception system which detects efficiently the road edges from the perception-domain image, and a fuzzy control system which uses the knowledge base information and the scene model to control the robot motion. On the other hand, the fuzzy control system is finely tuned through feed-backing mean square errors between the scene model parameters and the knowledge-base data. Hence, a road configuration from a preprocessed image is compared with a fuzzy template made from the fuzzy membership function based on the knowledge base module. Finally, the fuzzy controller uses results of this calculation to guide the robot on the planned path. This paper shows the principle of this system and the simulation results confirming the feasibility of the approach even in the presence of several image artifacts, such as sparse shadows and lighting changes..