Hardware Implementation of High-performance Classifiers for Edge Gateway of Smart Automobile

Abstract

Fog computing is a key solution for internet of things (IoT) applications, which demands operational security, real-time and power efficient intelligent responses, and low bandwidth usage. This paper introduces a novel idea related to an hardware implementation of High-performance classifiers for real-time and low power sensor data analytic on the intelligent edge gateway running on smart automobile. The high-performance classifiers uses an artificial neural network (ANN) to extract conclusive inferences from the raw automotive sensors information. The multiple classifiers are embedded into a re-configurable ANN hardware deign i.e. intellectual property core (IP core) which implemented and tested using field-programmable gate array fabric. In addition, this work studies the effect of the IP cores on the performance of the edge gateway. The implementation of fog/edge computing enables throughput reduction of 96.78% to 98.75% compared with the traditional gateway. The hardware design of the high-performance classifiers IP core requires only 31μ s and power consumption of 124mW for classification. The concept of re-configurable ANN model reduce about 41% to 93% of hardware resources requirement that contributing to reduced system power and cost.