Construction of an Interactive and Competitive Artificial Neural Network for the Solution of Path Planning Problems

摘要

In this paper we present an artificial visual system for autonomously moving vehicles. In this system visual information, extracted from the vehicle's surroundings by means of motion detection and range estimation strategies, is mapped onto a retinotopically ordered neural map. The activity pattern elicited in this neural map is projected onto an interactive competitive neural network which plans the next step of the vehicle on its path towards a given target location. The system has been implemented and tested for a two-dimensional, i.e. flat, world using computer simulations, giving encouraging results.