Driving quality assessment system as part of intelligent onboard system

Abstract

Driving parameters are controlled by external characteristics with the help of modern software tools; It can be, for example, the number of accelerations and deceleration, the speed of the vehicle, etc. However, the data obtained do not fully reflect the real situation. We propose a method of driving quality assessment by developing a neuro-fuzzy model of engine fuel consumption. In the first layer of the developed neuro-fuzzy model, the membership degrees of input parameters in fuzzy set are calculated. The outputs of the first layer neurons are the membership degrees of input values in fuzzy sets associated with the neurons. The second layer determines the degree to which the values of the input signals correspond to the rule conditions. The signals at the output of the third layer are the sum of products of weights and normalized degrees of the rules activity. It has been found that the cut of the surface appears smooth, which demonstrates that it is possible to obtain a control effect for any values of input variables from a given range. Values of the actual fuel consumption and fuel consumption according to the model are obtained. The relative error of individual measurements is calculated, as well as the average value of the model error. The maximum error was 3,5375 %, and the average model error was 0,4429 %. The developed method has been tested during road tests. The results of the tests confirm its adequacy on the public road (Ufa - Moscow route).