|
Abstract : |
In this paper, we address the following problem: given a robot which is at some unknown location in a known environment, how does the robot go about reaching its goal? In resolving this problem, we have abandoned the notion of localisation and instead developed the concept of a reactive network. A reactive network is simply a specialized kind of finite state machine whose states correspond to actions and whose transitions correspond to observations. We show that it is possible to generate reactive networks that can be used to reach a particular goal from any initial location, and hence to solve the problem of reaching a goal from an unknown location. In this paper, we develop the reactive network concept, describe how they can be constructed and use simulation to assess their performance. 1, |
