Spiking Neuron Model for Embedded Systems

Abstract

Spiking neural networks (SNN) are used in robotics, particularly on the boards of autonomous vehicles, so the issues related to the hardware implementation of spiking neurons and SNNs is hotly discussed. Significant attention is devoted to the energy efficiency of the models in use. In the frame of the presented project, well-established neuron models have been investigated. As the result the spikes counting model (SCM) enabling real-time operation and attaining high energy efficiency have been developed. The implementation of the developed model in microcontrollers MSP430 family is achieved without the need of floating-point operations (FPO). Moreover, we analyze the issue of transferring and implementing the spikes counting model using alternative platforms.