CC BY
6THE USE OF GNSS-MONITORING DATA IN THE SYNTHESIS OF VEHICLE DRIVING CYCLES
Manyashin A. V.
PhD in engineering Tyumen Industrial University Russian Federation, s. Tyumen, Volodarsky 38
Abstract
The car is the main polluter of the environment. This is due to the fact that until now most vehicles use a power plant powered by organic fuels. To determine the toxicity of the of exhaust gases, the engines are tested at transient modes - according to the so-called driving cycles. Driving cycles, as a rule, are a dependence of speed on current time, are widely used in road transport. The transient modes reproduced by driving cycles are predominant, especially in the use of cars in urban environments, so it is so important that the reference cycle used in estimating emissions corresponds to the actual operating conditions. The article considers the problem of establishing the characteristics of a typical driving cycle based on real high-speed profiles of cars obtained as a result of GNSS monitoring.
Keywords: driving cycles, satellite monitoring, trainable neural networks
In conditions of depletion of the world's reserves of organic fuels, the problem of energy saving in the operation of motor transport, which is the main consumer of the most valuable light petroleum products, is becoming more and more urgent. Fuel consumption by internal combustion engines is directly related to emissions of harmful substances during their operation. This forces specialists to work more actively on finding new technical solutions aimed at saving fuel and reducing the toxicity of exhaust gases during the operation of cars. Research is being conducted aimed at improving the design of cars and engines, automated traffic management, and optimization of traffic routes. In many cases, the standardized driving cycle of the car is taken as the basis of research. In order for the research results to be applicable in practice, it is necessary that the driving cycle used in the tests reflects the actual speed regime in the appropriate (typical) conditions. This is especially true of urban conditions. It is clear that even in the conditions of one country, the road situation in cities comparable in area and population density, the length of the road network, will differ significantly, not to mention the difference in the average speed and duration of its various phases (acceleration, steady-speed traffic, braking and stopping in megacities and small settlements). In these circumstances, finding a typical driving cycle that most adequately reflects the prevailing speed profiles of cars, for example, in urban conditions, is a difficult task.
In world practice, standardized, that is, officially approved for testing cars for fuel efficiency, speed profile templates are used. As a rule, such cycles are modal and non-modal. Among the first, the most famous are Japanese cycles 10-15 Mode and JC08, European NEDC, MNEDC, WLTP. The most well-known nonmodal ones include FTP-75 (USA) and Hyzem (EU). Modal driving cycles are characterized by a significant duration of phases with constant speed, while nonmodal ones consist almost entirely of transient modes (acceleration, acceleration) [1-3].
There are two main methods for obtaining driving cycles: direct recording of speed changes when driving on a representative road network and cycle synthesis based on processing statistical data on the speed profiles of cars in typical operating conditions, for example, in a city [4]. The first method is simpler, but has a
high error when using it as a standard. Synthesis of a typical cycle theoretically involves processing a huge array of data, and most importantly - time-consuming tests on test sections of the road network. The organization of such tests requires, in addition to specially equipped vehicles, compliance with a number of restrictions, for example, appropriate driver qualifications and regulation of the driving algorithm.
It is clear that obtaining synthetic typical driving cycles is a long process, and often the resulting pattern of speed changes no longer corresponds to the changed road conditions. For example, the still used European NEDC cycle was developed in the 80s of the last century.
Nevertheless, the need for adequate typical driving cycles obtained on the basis of statistical data for an acceptable time remains. To solve the problem of long-term experimental studies, it is possible to use up-to-date databases of satellite online monitoring, which has become widespread all over the world [5-6].
The object of research in this paper is the highspeed profile of vehicles. The subject is the velocity profile obtained by means of GNSS monitoring. The work uses a trained neural network of direct propagation.
As a statistical material for the synthesis of a driving cycle based on traffic speed reports generated by all servers of monitoring systems, only those where a sufficiently small update period is set are suitable. But even the speed profiles obtained with the minimum speed update frequency of one second at the moment require preprocessing. This is due to the fact that the velocity values are calculated by the GNSS system based on the coordinates obtained from the satellite and give a significant local error (Pic. 1), and it can be either positive or negative.
The figure shows jumps in the calculated speed values relative to the assumed real ones. Thus, smoothing of numerical values is necessary. To do this, you can use data coarsening using spectral data analysis tools, for example, Wavelet transform (Pic. 2). Approximation of GNSS monitoring data with a given detail allows you to smooth out outliers over the entire field of the velocity profile.
Pic. 1. The speed profile formed by the GNSS system
For multidimensional processing of the entire data the most "close" phases, and individual clusters will
array of real driving cycles, it is advisable to allocate serve as the basis for the design of the future synthetic
separate phases of movement with specified character- cycle. The final alignment of the resulting clusters - the
istics. This will make it possible to apply cluster analy- phases of motion - is carried out using Markov chains. sis in the future and actually determine the estimates of
Pic. 2. A velocity profile subjected to a Wavelet transform. Stamm 4.2
The identification of individual characteristic phases of motion and their characteristics in the analysed velocity profile is based on the theory of pattern recognition. However, the use of classification and identification methods is hampered by the "noise" of the data. Therefore, according to the author, it is advisable to use a trained neural network for this purpose. A direct propagation neural network can be designed and trained in the "Stamm" simulation program [7, 8] (Pic.
3). In order to be able to use the neural network to isolate individual phases of motion from the speed profile according to the GNSS monitoring report, it is necessary to perform its training on the reference speed profile, and its data must have the same rate fluctuations as the statistical material. To do this, you can use a typical GPS-GLONASS terminal connected to the provider's server, in the server settings you need to set the polling frequency equal to the interval of the available database of GNSS-monitoring speed profiles.
Pic. 3. Layout of the neural network in the program Stamm 4.2
In the process of forming a training sample, it is necessary to set parameters that will then be used in the development of a program for recognizing individual phases of movement. One of the main parameters is the position in the source file of the speed profile, where the selected phase of movement ends. This position serves as a pointer from which the recognition of the next phase begins.
By analysing the array of velocity profiles obtained by GNSS monitoring, it is necessary to determine the maximum number of neural network inputs. It should be large enough to completely cover all points belonging to the longest phase of the movement.
The output of the developed motion phase recognition program should be a text file, each line of which contains the parameters of a separate phase. The resulting file can later be subjected to cluster analysis in the "Stamm" program to determine the parameters of clusters representing similar phases of movement and their number. At the same time, in the cluster analysis settings, to automatically determine the number of clusters across the entire sample, you must specify "0" in the "Number of clusters" field of the settings panel.
The proposed methods and approaches to the organization of the experiment will significantly reduce the amount of experimental research in the synthesis of typical driving cycles of vehicles. This, in turn, will significantly reduce the time of their receipt, which means that it will increase the adequacy of reproducing real operating modes with standardized speed profiles.
References
1. Manyashin, A.V., Manyashin, S.A., Modelling of fuel consumption by cars based on typical driving cycles, TyunGNGU: Tyumen, 2014 - 124 p. [Published in Russian].
2. United States Environmental Protection Agency. (1993). Federal Test Procedure Review Project. Status Report 420-R-93-006. Washington, Office of Air and Radiation, February, 1993, 21 p.
3. Anida, I. N. & Rahman, S. A. (2019). Driving cycle development for Kuala Terengganu city using k-means method. International Journal of Electrical and Computer Engineering, 9(3), pp. 1780-1787. DOI: 10.11591/ijece.v9i3.pp1780-1787.
4. Manyashin, A.V., Manyashin, S.A., Method for synthesizing the driving cycle of a car. International scientific journal, 1, pp. 87-91, 2012 1. [Published in Russian].
5. Covaciu, D., Preda, I., Florea, D., & Vasile C. (2010). Development of a driving cycle for Brasov city. International Conference of Mechanical Engineering ICOME, October, 2010. Craiova, Universitaria Publishing House, T. II, pp. 761-766.
6. Lipar, P., Strnad, I., Cesnik, M., & Maher, T. (2016). Development of Urban Driving Cycle with GPS Data Post Processing. Promet-Traffic&Transportation, 28(4), pp. 353-364. DOI: 10.7307/ptt.v28i4.1916
7. Manyashin, A.V., Using Stamm 3.0 for solving scientific and engineering problems, TIU: Tyumen, 2017 - 191 p 1. [Published in Russian].
8. Manyashin, A.V., Statistical data analysis and simulation in the system Stamm 4.0, TIU: Tyumen, 2020 - 216 p 1. [Published in Russian].
