Making Cognitive Map of Outdoor Environment

Abstract

Exploring an unknown environment and making its model is a key issue for realizing an intelligent robot . This paper presents a framework for representing an outdoor environment along a rout e as a Cognitive Map. Continuous views by a moving robot are transformed into a pathcentered 2(1/2)D scene rep resentation, ca lled Panoramic Repres e ntation, which the cognitive mapping process tak es as input . Range databased analysis of the Panoramzc R epresen tation , segments th e viewed scene into ground and each object standing upon it. The objects are then arranged into a structure map which provides robust qualitativ e information on th e environment structure. Although rang e data in the 2( 1/2)D re presentation are non-prec is e and vary with th e camera speed and orientation, the spatial relations and relative sizes of the objects are invariant. The Cognitive Map is desc ribed qualitatively in terms of places along the rout e and landmarks found as distinctive objects in eac h place . When the robot mov es along the same rout e, the Cognitive Map guides the navigation; the robot can reliably match the memorized landmarks against objects appearing in its sight.