\Meaning" through Clustering by Self-Organisation of Spatial and Temporal Information

This paper presents an episodic mapping mechanism used for the self-localisation of autonomous mobile robots. A two layer self organising neural network classiies perceptual and episodic information to identify \perceptual landmarks" (and thus the robot's position in the world) uniquely. Through this process relevant information is obtained from the temporal ow of ambiguous and redundant sensory information, such that meaningful internal representations of the robot's environment emerge through an unsupervised process of self-organisation, constructing an analogy to the real world.