Spiking Neural Network for Enhanced Mobile Robots’ Navigation Control

Abstract

Contemporary robotics primarily emphasizes autonomous mobile robots, and Artificial Neural Networks (ANNs) have demonstrated their proficiency in managing intricate, nonlinear systems with illusive models. This study explores the progress made in third-generation neural networks, namely Spiking Neural Networks (SNNs), which has capabilities that beyond those of traditional NNs. We introduce a modular mobile robot navigation controller that utilizes SNNs to transmit both spatial and temporal information. The controller is constructed and evaluated within a simulated environment that accurately simulates real-life situations, utilizing promising Spiking Neural Networks (SNNs). This study seeks to improve the autonomous robot’s collision avoidance and navigation capabilities by implementing a three-layered spiking neural network (SNN). Utilizing a customized variant of Spike-Timing-Dependent Plasticity (STDP) to train inhibitory synapses enhances the network’s efficiency, resulting in a reduced number of required training iterations. A potential method for identifying synaptic connections in a concealed-layer SNN is introduced as an evolutionary methodology, which might be utilized to establish synaptic connectivity in the paper.