To understand the behaviour of natural autonomous systems,research is carried out on artificial autonomous agents. The paperfocuses on how simple behaviours can be learnt autonomously using abootstrapping method. Firstly, a two dimensional self-organising map isrealised which provides the agent's sense of orientation. Once thisrelative positioning system has been established, the agent learns tonavigate towards a target using the reinforcement learning technique ofQ-learning. Since only neural network processing is used, this techniqueemulates the distributed and adaptive information processing found innatural autonomous systems. Furthermore, due to its generality, theneural implementation developed is transferable to other artificialautonomous agents with different sensors and effector suites
展开▼
