Development of Vision‐Based Navigation System for Wheeled Agent

Abstract

AbstractIn this paper, a vision‐based navigation system including support vector machine (SVM) path planner and fuzzy sliding‐mode controlled (SMC) path follower is developed for a wheeled agent. The developed system, comprising image acquisition, map formatting, path planning, label assignment, and path tracking inference mechanisms, aims to gracefully follow a planned smooth path. To obtain accurate coordinate values, the captured images are first adjusted by calibration processing. As a roadmap method for path generation, the Voronoi diagram is employed as a preprocessor and the Gaussian kernel SVM postprocessor is applied consecutively. To deal with the uncertainties, a path follower based on fuzzy SMC is embedded to track the planned path on line. In this study, a practical framework is implemented to assess the performance. With the real devised system, a series of experiments are carried out and analyzed to confirm the expected performance. The experiments show a robust capability of the system for both path planning and path tracking under various obstacle layouts.