Modelling and forecasting of car speed using detrended inputs and hybrid multilayered percepton network

Abstract

The study explores the capability of hybrid multilayered perceptron network (HMLP) trained using modified recursive prediction error (MRPE) algorithm as a nonlinear model for a speed of a car. This model is used for speed forecasting for a car in order to avoid speed limit on the road traffic. The injected fuel was measured from the experimented car and used as an input for the HMLP model. The input and output variables were recorded from the car sensors. The modeling performances are measured based on the correlation tests. The simulation result shows that preprocessing data using digital high-pass filter is better than the first differential to pass the correlation test. This preliminary result shows that this modeling system is capable to be used as an alternative speed-forecaster for cars.