Estimation of Driving Resistance Coefficients with Neural Network and Its Low-Power Implementation using FPGA

Abstract

It is important to acquire the states of a car not only on test environments but also on real roads to realize a carbonneutral society. However, it is labor- and cost- inefficient due to the need of many in-car sensors. To alleviate this issue, this research acquires driving resistance coefficients by not directly measuring them but by predicting them using a neural network. In addition, we implement the proposed neural network model on an FPGA to enable its execution within the spare power of a car.