The Application of Spiking Neural Networks in Autonomous Robot Control

Abstract

Artificial neural networks have a wide range of applications nowadays in which they are used for intelligent information processing. This paper deals with an application of spiking neural networks in autonomous mobile robot control. The topology of the implemented spiking neural networks was developed through a modified genetic algorithm and through the process of autonomous interaction with the scene environment. Since the genetic algorithm did not use a crossover operator we adapted the mutation operator adding a constraint that prevented creation of a new generation of population with weak individuals in comparison with the previous generation of population. The paper proposes a parallel combination of both left and right local spiking neural network as well as a practical implementation of this proposition in the form of an intelligent navigation system in an autonomous mobile robot. This design enhances the implemented navigation system with a new cognitive property of intelligent information processing using a spiking neural network. Having been adapted to the scene environment, the navigation system was able to make right decisions, change its direction and refrain from collision with the scene walls.