Context-aware robot navigation based on sensor association rules

Abstract

Within the mobile robotics research community, a great many approaches have been proposed for solving the navigation problem. The key difference between these various navigation architectures is the manner in which they decompose the problem into smaller subunits. In this paper, a data mining methodology developed for the retrieving significant frequent patterns is extended to allow robots to learn and navigate on unknown terrain in natural way. The method has two phases: context identification phase and validation phase. Conjunction of those phases provides an easy and straightforward way for exploring new workings space for robots.