AN INNOVATIVE NEURAL NETWORK BASED ON THE TOAD'S VISUAL SYSTEM

Abstract

This paper considers the development of an artificial neural network based on the visual response of toads to robotics applications. The network structure combines elements of both modular and fully connected networks. The network is remarkable at analyzing the ‘What’ and ‘Where’ aspects of a visual scene. This results in ability to process individual segments of information in a complex visual environment. These attributes make the network particularly suitable for obstacle avoidance in robotics. Results are presented which show these qualities of the network in simplified circumstances. The network used is trained using a genetic algorithm and reinforcement based reward scheme.