Autonomous vehicle interaction with in-door environments

Abstract

The paper describes the possibilities of enhancing the performance of autonomous vehicles by interacting with the surrounding environment. Benefits and drawbacks of the use of different sensors and strategies for integrating the acquired knowledge are examined in an example case of a ‘through partly closed door’ navigation task for an AGV. This illustrates the increased ability of sensors to acquire knowledge by actively doing explorative actions. A combined mechanical actuation and probing device accentuates the sense-act paradigm.

The theory and algorithms used are described in detail, and extensive experiments with an implementation of the ‘through door’ navigation algorithm on our non-holonomic test-bed Autonomous Guided Vehicle are documented.