Scene Memory of Mobile Robot based on Competitively Growing Neural Network using Temporal Coding

This paper describes the saliency-based scene memory model of a mobile robot in which objects in salient spots are quickly learned and recognized based on the competitively growing neural network using temporal coding. This neural network represents objects using latency-based temporal coding and grows size and recognizability through self-organized learning with growth. In this model, objects in saliency-based attended spots are sequentially encoded to be invariant with respect to position and size by this neural network and their positions and sizes are encoded simultaneously. Through experiments using a Khepera robot equipped with a camera, it is shown that quick self-organized learning and glance recognition of objects in scenes are well performed by our model.