Visual mobile robots perception for motion control

Abstract

Visual perception methods are developed first mainly for human perception description and understanding. The results of these researches are now very popular for robots visual perception modeling. In this chapter is present first a brief review of the basic visual perception methods suitable for intelligent mobile robots applications. The analysis of these methods is directed to the mobile robot motion control, where the visual perception is used for objects or human body localization like: Bayesian visual perception methods for localization; log-polar visual perception; area of robot observation mapping using visual perception; landmark-based finding and localization with visual perception etc. The development of an algorithm for mobile robot visual perception is proposed based on the features of log-polar transformation to represent some of the objects and scene fragments in area of mobile robot area of observation in a more simple form for the image processing. The features and advantages of the proposed algorithm are demonstrated with the popular for the mobile robots visual perception situation of motion control in a road or corridor with outdoor road edges, painted lane separation lines or indoor two side existing room or corridor lines. The proposed algorithm is tested with suitable simulations and the experiments with real mobile robots like Pioneer 3-DX (Mobil Robots INC), WiFiBot and Lego Robot Mindstorms NXT. The results are summarized and presented as graphics, test images and comparative tables in the conclusion. © Springer-Verlag Berlin Heidelberg 2012.