Navigation of intelligent vehicle using visual information. Experimental Evaluation of Hierarchical controller Divided by Information Type.

Abstract

We aim at developing an intelligent machine by taking up an intelligent vehicle that is to get to a destination without touching walls on corridors in a building using visual inforration. Robustness of systems and facility of designing are considered important on the ground that it is indispensable to achieve intelligent machines in real world. Therefore, we adopt a hierarchical architecture which has a NAV layer described as mode representation including map information and DRV layer made up of some behavior-based-modules. Moreover, interpreting visual information qualitatively avoiding excessively quantitative interpretation to recognize characteristic places allows the system to get to the destination without artificial landmarks. An experiment with a vehicle which is fully equipped with camera devices, computing devices, battery, and driving system showed the effectiveness of this method.