Capítulo 4: Desarrollo de la aplicación móvil
4.2. Esquema de la aplicación
This research firstly analysed real traffic data gathered by Brisbane City Council exploiting Bluetooth technology. This means that by dividing urban arterials into small sections between each continuous couple of installed Bluetooth detectors and matching the same captured devices from downstream and upstream, travel time for each individual vehicle on that section is calculated. Then, after identifying and removing outliers from the data, travel time series for whole corridor is estimated and saved in databases for further calculations. This data is clustered into different groups of days of week, working day, holiday and school holiday. Then daily travel time series from the same clusters are taken out from databases later to use in prediction models. Therefore, based on the reviewed literature and characteristics of data, the Seasonal Auto Regressive Integrated Moving Average model is created, calibrated on data and applied to forecast future travel time values for some random days. This model is used to forecast future travel time series, both with consideration of the seasonality factor in historical data and also without consideration. The seasonal model and non-seasonal one are tested on the same databases and travel time on some random days from three different arterial corridors with a variety of traffic conditions is predicted. Based on calculating error rate associated with each prediction, two models are compared and both are evaluated.
5.1 BLUETOOH DATA RELIABILITY
In terms of Bluetooth data reliability for estimating travel time on urban arterials, first of all a target of required 90 percent accuracy was outlined for sampling data. For this purpose a minimum required number of data points are defined for each sample window for estimating each minute’s travel time value. However, after creating sample size for estimating travel time on each minute of the day, it is found that the 90 percent confidence target is not met by available Bluetooth data in majority of cases. Based on comparing available data points with the required number of data points within each sample window, for some corridors error of estimation even reached to 70% during the peak periods. This means that some gaps are observed during the peak times in daily travel time data. Therefore,
based on the results of this study, the current rate of observed Bluetooth devices is not accurate enough for estimating travel time on Brisbane urban arterial corridors. Average confidence level for travel time estimation by available data for three case study routes is defined between 44 to 85 percent in this research. Depending on the required estimation accuracy in traffic studies it should be defined whether Blutooth data are accurate enough to be used as the main data source or not.
5.2 SHORT TERM TRAVEL TIME PREDICTION
After Bluetooth data reliability investigation in this study, it is assumed that the travel time series estimated by using Bluetooth data are accurate and represent the real travel time values on the selected study routes. Accordingly, the next steps of travel time forecasting are done on the created database from Bluetooth data.
For predicting short term travel time on urban arterials, separate databases are created for each route. According to the explained defined clusters in analyses, historical data for predicting each day is chosen and the SARIMA model considering seasonality and without considering it is applied on the data. Finally, travel time series for several random days on each of the three case study corridors are predicted within different prediction horizons.
Based on the results taken from all three case study routes with different traffic conditions, the SARIMA model considering seasonality in travel time series’ historical data performs well up to 30 minutes after real data. This statement is driven by comparing the error rate of prediction by SARIMA to the error rate of the historical mean method. This comparison showed that the error of predicting travel time series by SARIMA increases by increasing the prediction horizon. But as the error of prediction gets more than the historical mean method, it is not acceptable to use SARIMA for predicting travel time more than 30 minutes after real data. All these conclusions are focused on congestion hours, which are more critical to be predicted in travel time series.
Testing the model without considering the seasonality factor in predictions turned the model to an ARIMA model. So by running the suitable ARIMA model on the same databases, the associated error for prediction by ARIMA is calculated as well. In this case prediction error by ARIMA has increased dramatically due to the
lag between real and predicted travel time values. This means that application of the ARIMA model for predicting travel time on urban arterials is not approved.
5.3 SARIMA MODEL RELIABILTY
As the main objective of this research for predicting future travel time values on arterials, reliability of the SARIMA model is also evaluated under different conditions. As the model is applied on different routes with a variety of travel time conditions, the 90th percentile of error for each prediction horizon’s predictions is calculated. Therefore, based on calculated 90th percentile of error for each prediction horizon, travel time prediction by the SARIMA model in at least 90% of the cases concludes more accurate results compared to the historical mean method with a considerable rate of difference in errors. The 90th percentile value for peak periods is calculated between 10 to 40 percent for 5 to 30 minutes prediction horizon, which is still less than 50 percent error of historical average method’s results. This means that in the worst case for 30 minutes prediction horizon, 90 percent of results by SARIMA have less error than 40 percent. The associated error within historical mean method is calculated as 50 percent. Accordingly SARIMA predict short term travel time better than historical mean method for up to 30 minutes after real data.
Finally, depending on the required application of the SARIMA model on travel time series - prediction or regression -considering seasonality is necessary for online travel time prediction (having no data after the current time) on routes. In terms of using the model for regression purposes such as finding missing data in travel time series in offline mode, there is no need to consider seasonality in calculations. Generally, according to the discussed results in this thesis, both these two models can show good performance in terms of time series prediction or regression methods, based on recurrent behaviour in historical databases.
5.4 FURTHER FUTURE RESEARCH
Consequently, it is concluded that the Seasonal Auto Regressive Integrated Moving Average model could potentially be applied as a linear model on historical databases for predicting short-term travel time on arterial networks. However, future studies would be needed for improving the accuracy of prediction results via consideration of incidental conditions on arterials. There would also be good potential for further study in terms of increasing the confidence of corridor travel
time estimation by improving the rates of travel time data points. In this case, fusing data from a multisource such as Bluetooth and loops is recommended.
Appendices
Sample SARIMA code for forecasting short term travel time within different prediction horizons:
%function SARIMA_Model(HDN,MonthDay)
% This function fit a SARIMA(1,0,1) model on the data and the forecast
% future TT values for each 5 minutes in different prediction horizons)
% HDN= Historical Data points Number = the number of data points save in
% the historical data base file. for example the file "Whole_Data" in
% "Jun7_Data_TW" file contains 7592 data point. % Default calling syntax:SARIMA_Model(7592,'Jun7')
HDN =1566;
model =
arima('D',0,'Seasonality',287,'ARLags',1,'SARLags',287,'MALags',1,'S
MALags',287)
model = arima('D',0,'ARLags',1,'MALags',1)
% infilename = [MonthDay '_Data_TW.mat']; % load(infilename);
fit=estimate(model, Jun4_peaks_data_TW);
Outfilename = [MonthDay '_forecasted101.mat'];
%%%%
%%%%% forecasting by fttied model:
%%%%%%%% for 5 mins forecast:
Forecasted5=[];
for i=0:1:87
[Yf YMSE] = forecast(fit,1,'Y0',final_peaks_jun4(1:HDN+i,1));
Forecasted5=[Forecasted5;Yf]; end
%%%%%%%% for 15 mins forecast:
Forecasted15=[];
for i=0:1:87
[Yf YMSE] = forecast(fit,3,'Y0',final_peaks_jun4(1:HDN+i,1));
Forecasted15=[Forecasted15;Yf(3,1)];
end
%%%%%%%% for 30 mins forecast
Forecasted30=[];
for i=0:1:87
[Yf YMSE] = forecast(fit,6,'Y0',final_peaks_jun4(1:HDN+i,1));
Forecasted30=[Forecasted30;Yf(6,1)];
end
%%%%%%%% for 75 mins forecast
Forecasted75=[];
for i=0:1:87
[Yf YMSE] = forecast(fit,9,'Y0',final_peaks_jun4(1:HDN+i,1));
Forecasted75=[Forecasted75;Yf(9,1)];
end
%%%%%%%%% for 70 mins forecast
Forecasted60=[];
for i=0:1:87
[Yf YMSE] = forecast(fit,12,'Y0',final_peaks_jun4(1:HDN+i,1));
Forecasted60=[Forecasted60;Yf(12,1)]; end Forecasted90=[];
for i=0:1:87
[Yf YMSE] = forecast(fit,18,'Y0',final_peaks_jun4(1:HDN+i,1));
Forecasted90=[Forecasted90;Yf(18,1)]; end save(Outfilename); % end
Figure 20: First section daily travel time profiles for August 2011 in Coronation Drive
2014 Amir Mohammad Khoei
Bibliography
Abbott-Jard, M., H. Shah and A. Bhaskar. eds. 2013. Empirical evaluation of bluetooth and wifi scanning for road transport 36th Australasian Transport
Research Forum (ATRF), Brisbane, Australia, 2 - 4 October. Brisbane,
Australia.
Abedi, N., A. Bhaskar and E. Chung. eds. 2013. Bluetooth and wi-fi mac address based crowd data collection and monitoring: Benefits, challenges and enhancement. 36th Australasian Transport Research Forum (ATRF),
Brisbane, Australia, 2 - 4 October. Brisbane, Australia.
Abedi, N., A. Bhaskar and E. Chung. 2014. Tracking spatio-temporal movement of human in terms of space utilization using media-access-control address data.
Applied Geography 51 (0):72-81.
http://www.sciencedirect.com/science/article/pii/S0143622814000629. AIDOO, E. 2011. Forecast performance between sarima and setar models: An
application to ghana inflation rate, Uppsala University.
Anderson, O. D. 1976. Time series analysis and forecasting: The box-jenkins
approach. Butterworths London, UK.
Araghi, B. N., R. Krishnan and H. Lahrmann. Application of bluetooth technology for mode-specific travel time estimation on arterial roads: Potentials and challenges.
Asudegi, M. 2009. Optimal number and location of bluetooth sensors for travel time data collection in networks, University of Maryland (College Park, Md.).
http://drum.lib.umd.edu/bitstream/1903/9400/1/Asudegi_umd_0117N_10478. pdf.
Bajwa, S. I., E. Chung and M. Kuwahara. eds. 2003. A travel time prediction method based on pattern matching technique 21st ARRB and 11th REAAA
Conference, Cairns, Australia,. Cairns, Australia,.
Bakula, C., W. H. Schneider, Iv and J. Roth. 2012. Probabilistic model based on the effective range and vehicle speed to determine bluetooth mac address matches from roadside traffic monitoring. Journal of Transportation
Engineering 138 (1):43-49. http://dx.doi.org/10.1061/(ASCE)TE.1943- 5436.0000284.
Barceló, J., L. Montero, L. Marqués and C. Carmona. 2010. Travel time forecasting and dynamic origin-destination estimation for freeways based on bluetooth
traffic monitoring. Transportation Research Record: Journal of the
Transportation Research Board 2175 (-1):19-27.
http://dx.doi.org/10.3141/2175-03.
Bhaskar, A. 2009. A methodology (cuprite) for urban network travel time estimation by integrating multisource data, School of Architecture, Civil and Environmental Engineering, Ph.D. Thesis ,The Ecole Polytechnique Fédérale de Lausanne (EPFL)
Lausanne. http://library.epfl.ch/theses/?nr=4416.
Bhaskar, A. and E. Chung. 2013a. Fundamental understanding on the use of bluetooth scanner as a complementary transport data. Transportation
Research Part C: Emerging Technologies 37:42-72.
Bhaskar, A. and E. Chung. 2013b. Fundamental understanding on the use of bluetooth scanner as a complementary transport data. Transportation
Research Part C: Emerging Technologies 37 (0):42-72.
http://www.sciencedirect.com/science/article/pii/S0968090X13002027. Bhaskar, A., E. Chung and A.-G. Dumont. 2008. Benefits and issues of urban
network travel time prediction. In 7th European Congress and Exhibition on
Intelligent Transportation Systems and Services, Geneva, Switzerland, 4-6 June 2008.
Bhaskar, A., E. Chung and A.-G. Dumont. 2009. Estimation of travel time on urban networks with midlink sources and sinks. Transportation Research Record:
Journal of the Transportation Research Board 2121:41-54.
http://dx.doi.org/10.3141/2121-05.
Bhaskar, A., E. Chung and A.-G. Dumont. 2010. Analysis for the use of cumulative plots for travel time estimation on signalized network. International Journal
of Intelligent Transportation Systems Research 8 (3):151-163.
http://dx.doi.org/10.1007/s13177-010-0012-y.
Bhaskar, A., E. Chung and A.-G. Dumont. 2011. Fusing loop detector and probe vehicle data to estimate travel time statistics on signalized urban networks.
Computer-Aided Civil and Infrastructure Engineering 26 (6):433-450.
http://dx.doi.org/10.1111/j.1467-8667.2010.00697.x.
Bhaskar, A., E. Chung and A.-G. Dumont. 2012. Average travel time estimations for urban routes that consider exit turning movements. Transportation Research
Record: Journal of the Transportation Research Board 2308 (-1):47-60.
http://dx.doi.org/10.3141/2308-06.
Bhaskar, A., E. Chung, M. Kuwahara, O. De Mouzon and A.-G. Dumont. 2007. Urban network travel time estimation from stop-line detector data and signal controller data. In Wit transactions on the built environment, ed. C. A.
Brebbia, 147-156. Coimbra, Portugal WIT Press.
http://www.scopus.com/scopus/inward/record.url?eid=2-s2.0- 38849164218&partnerID=40&rel=R8.0.0
Bhaskar, A., L. M. Kieu, M. Qu, A. Nantes, M. Miska and E. Chung. 2014. Is bus overrepresented in bluetooth mac scanner data? Is mac-id really unique?
International Journal of Intelligent Transportation Systems Research (http://dx.doi.org/10.1007/s13177-014-0089-9 ).
Bhaskar, A., K. Le Minh, M. QU, A. Nantes, M. Marc and E. Chung. eds. 2013. On the use of bluetooth mac scanners for live reporting of the transport network.
The 10th International Conference of Eastern Asia Society for Transportation Studies, Taipei, Taiwan, 9-12 September. Taipei, Taiwan.
Bhaskar, A., M. Qu and E. Chung. 2014a. Bluetooth vehicle trajectories by fusing bluetooth and loops: Motorway travel time statistics IEEE Transactions on
Intelligent Transportation Systems (in press, DOI: 10.1109/TITS.2014.2328373).
Bhaskar, A., M. Qu and E. Chung. 2014b. A hybrid model for motorway travel time estimation- considering increased detector spacing. Transportation Research
Record: Journal of the Transportation Research Board (in press).
Blogg, M., C. Semler, M. Hingorani and R. Troutbeck. 2010-9. Travel time and
origin-destination data collection using bluetooth mac address readers.
Australasian Transport Research Forum 2010 Proceedings
Bluetooth: About the technology. 2011. http://www.bluetooth.com/Pages/About-the- Technology.aspx.
Bowerman, B. L., R. T. O'Connell and T. Richard. 1993. Forecasting and time
series: An applied approach. Belmont CA Wadsworth.
Box, G. 1976. P, and jenkins, gm,“time series analysis: Forecasting and control,”.
Time Series and Digital Processing.
Brennan, J. T. M., J. M. Ernst, C. M. Day, D. M. Bullock, J. V. Krogmeier and M. Martchouk. 2010. Influence of vertical sensor placement on data collection efficiency from bluetooth mac address collection devices. Journal of
Transportation Engineering 136 (12):1104-1109.
http://dx.doi.org/10.1061/(ASCE)TE.1943-5436.0000178.
Carpenter, C., M. Fowler and T. J. Adler. 2010. Generating route specific origin- destination tables using bluetooth technology. In TRB 2012 Annual Meeting.
Casey, B., A. Bhaskar and E. Chung. 2012. Data requirements and graph data structures for a multi-modal, multi-objective trip planner. In 25th Australian
Road Research Board Conference, Perth, Australia.
Casey, B., A. Bhaskar, H. Guo and E. Chung. 2014. Critical review of time
dependent shortest path algorithms, an intermodal trip planner perspective.
Transport Reviews: A Transnational Transdisciplinary Journal DOI:
10.1080/01441647.2014.921797.
Click, S. M. and T. Lloyd. 2012. Applicability of bluetooth data collection methods for collecting traffic operations data on rural freeways. In TRB 2012 Annual
Meeting: Department of Civil and Environmental Engineering Tennessee
Tech University.
D., D. and P. a. M. J. Vickich. 2010. Bluetooth-based travel time/speed measuring
system development University Transportation Center for Mobility Texas
Transportation Institute.
Gibson, D. 2009. How loop detectors work.
Haghani, A. and Y. Aliari. 2011. Using bluetooth sensor data for ground-truth testing of reported travel times. In TRB 2012 Annual Meeting: Department of Civil and Environmental Engineering University of Maryland, College Park. Haghani, A., M. Hamedi, K. F. Sadabadi, S. Young and P. Tarnoff. 2010. Data
collection of freeway travel time ground truth with bluetooth sensors.
Transportation Research Record (2160):60-68.
http://dx.doi.org/10.3141/2160-07.
Hainen, A., J. Wasson, S. Hubbard, S. Remias, G. Farnsworth and D. Bullock. 2011. Estimating route choice and travel time reliability with field observations of bluetooth probe vehicles. Transportation Research Record: Journal of the
Transportation Research Board 2256 (-1):43-50.
http://dx.doi.org/10.3141/2256-06.
Haseman, R., J. Wasson and D. Bullock. 2010. Real-time measurement of travel time delay in work zones and evaluation metrics using bluetooth probe tracking.
Transportation Research Record: Journal of the Transportation Research Board 2169 (-1):40-53. http://dx.doi.org/10.3141/2169-05.
Israel, G. D. 1992. Determining sample size. University of Florida Cooperative Extension Service, Institute of Food and Agriculture Sciences, EDIS. J.Barceló, L.Montero, M.Bullejos, O. Serch and C. Carmona. 2011. A kalman filter
approach for the estimation of time dependent od matrices exploiting
Statistics and Operations Research and CENIT (Center for Innovation in Transport) Technical University of Catalonia.
Khoei, A. M., A. Bhaskar and E. Chung. 2013. Travel time prediction on signalised urban arterials by applying sarima modelling on bluetooth data. In
Australasian Transport Research Forum 2013.
Kieu, L. M., A. Bhaskar and E. Chung. 2012. Bus and car travel time on urban networks: Integrating bluetooth and bus vehicle identification data. Kieu, L. M., A. Bhaskar and E. Chung. eds. 2013. Empirical evaluation of public
transport travel time variability. 36th Australasian Transport Research
Forum (ATRF), Brisbane, Australia, 2 - 4 October. Brisbane, Australia.
Kristin Tufte, S. K. 2008. Assessment and refinement of real-time travel time
algorithms for use in practice: Oregon Transportation Research and
Education Consortium (OTREC).
Maier, H. R. and G. C. Dandy. 1996. Neural network models for forecasting univariate time series. Neural Network World 5:747-772.
Makridakis, S., S. C. Wheelwright and R. J. Hyndman. 2008. Forecasting methods
and applications. Wiley. com.
Malinovskiy, Y. and Y. Wang. 2011. Pedestrian travel pattern discovery using mobile bluetooth sensors. In TRB 2012 Annual Meeting: Department of Civil and Environmental Engineeringn University of Washington.
MALNATI, G., A. CARLEVATO, G. PORTELLI and C. BARBERIS. Bluetown:
Extracting floating ttarnspot data from personal mobile devices via bluetooth.
Martchouk, M. and F. Mannering. 2009. Analysis of travel time reliability on indiana
interstates: USDOT Region V Regional University Transportation Center.
Martchouk, M., F. Mannering and D. Bullock. 2011. Analysis of freeway travel time variability using bluetooth detection. Journal of Transportation Engineering 137 (10):697-704. http://link.aip.org/link/?QTE/137/697/1
http://dx.doi.org/10.1061/(ASCE)TE.1943-5436.0000253.
McKenzie, E. 1984. General exponential smoothing and the equivalent arma process.
Mei, Z., D. Wang and J. Chen. 2012. Investigation with bluetooth sensors of bicycle travel time estimation on a short corridor. International Journal of
Distributed Sensor Networks 2012. http://dx.doi.org/10.1155/2012/303521. Murphy, P., E. Welsh and J. P. Frantz. eds. 2002. Using bluetooth for short-term ad
hoc connections between moving vehicles: A feasibility study. Vehicular
Technology Conference, 2002. VTC Spring 2002. IEEE 55th: IEEE.
Nantes, A., M. Miska, A. Bhaskar and E. Chung. 2014. Noisy bluetooth traffic data?
Road & Transport Research 23 (1):33-43.
Nusser, R. and R. M. Pelz. eds. 2000. Bluetooth-based wireless connectivity in an automotive environment. Vehicular Technology Conference, 2000. IEEE
VTS-Fall VTC 2000. 52nd: IEEE.
O’Neill, E., V. Kostakos, T. Kindberg, A. Schiek, A. Penn, D. Fraser and T. Jones. 2006. Instrumenting the city: Developing methods for observing and understanding the digital cityscape eds. P. Dourish and A. Friday, 315-332: Springer Berlin / Heidelberg. http://dx.doi.org/10.1007/11853565_19. Okutani, I. and Y. J. Stephanedes. 1984. Dynamic prediction of traffic volume
through kalman filtering theory. Transportation Research Part B:
Methodological 18 (1):1-11.
Ong, C.-S., J.-J. Huang and G.-H. Tzeng. 2005. Model identification of arima family using genetic algorithms. Applied Mathematics and Computation 164