A TRAVEL TIME ESTIMATION MODEL OF PRIVATE CARS IN URBAN ARTERIAL BASED ON HETEROGENEOUS TRAFFIC

Transcription

1 International Journal of Civil Engineering and Technology (IJCIET) Volume 8, Issue 7, July 2017, pp , Article ID: IJCIET_08_07_073 Available online at aeme.com/ijciet/issues.asp?jtype=ijciet&vtyp pe=8&itype=7 ISSN Print: and ISSN Online: IAEME Publication Scopus Indexed A TRAVEL TIME ESTIMATION MODEL OF PRIVATE CARS IN URBAN ARTERIAL ROADS BASED ON HETEROGENEOUS TRAFFIC Sahrullah Doctoral Student, Hasanuddin University, Faculty of Engineering, Department of Civil Engineering, South Sulawesi, Indonesia Muh. Isran Ramli Associate Professor, Hasanuddin University, Faculty of Engineering, Department of Civil Engineering, South Sulawesi, Indonesia Nur Ali Associate Professor, Hasanuddin University, Faculty of Engineering, Department of Civil Engineering, South Sulawesi, Indonesia Ramli Rahim Professor, Hasanuddin University, Faculty of Engineering, Department of Architecture, South Sulawesi, Indonesia ABSTRACT Heterogenous traffic condition on road network in many urban roads in developing countries leads to uncertainty travel time for motor vehicles, especially for private cars. In this regard, the present study aims to model the travel time of the private cars on arterial urban roads under heterogeneous traffic situation. The study focused on the urban road network in Makassar City, Indonesia as a case study. The travel time model was developed using a linear regression model approach which based on dynamic motion characteristics of the private cars. The travel time data collection of the private cars utilized a GPS instrument, where twenty arterial urban roads in the city were selected as the survey location. The test vehicle of the private car for travel time survey was selected the car type which majority in traffic composition in the roads. Regarding the floating car methods, the travel time and dynamic speed in second by second of the test vehicle was measured. The measurement involved three peak hour periods of the traffic condition. Regarding the dynamic speed of the vehicle, the characteristics of the dynamic vehicle motion such as acceleration, deceleration, idle time, crushing time, and average speed were obtained. Further, the relationship models between travel time and the characteristics of the dynamic vehicle motion were developed based multiple linear regression approach. The modelling results showed that cruising time, acceleration, and deceleration of the private car motion were significant variables in the travel time editor@iaeme.com

2 Sahrullah, Muh. Isran Ramli, Nur Ali and Ramli Rahim model of the private cars in arterial urban roads under heterogeneous traffic situation. Key words: Travel time, private car, urban arterial road, heterogeneous traffic. Cite this Article: Sahrullah, Muh. Isran Ramli, Nur Ali and Ramli Rahim. A Travel Time Estimation Model of Private Cars in Urban Arterial Roads Based on Heterogeneous Traffic. International Journal of Civil Engineering and Technology, 8(7), 2017, pp INTRODUCTION Nowadays, many cities in developing countries such cities in Indonesia are facing serious traffic problems due to the motor vehicles growth has increased rapidly. In this situation, the vehicles made maneuvers and behaviors that are insufficient for the condition. In example, light vehicles and motorcycles have conducting zigzag maneuvers, creeps up slowly to the front of queue when the signal are red, impedes traffic flow by disturbing the star of other vehicle behind, etc. (Chandra et al, 2003; Zakaria et al. 2011, Hustim et al., 2011). Also, the vehicles have inconsistency or indiscipline to use their lane (Aly et al., 2011, 2012; and Hustim et al., 2011). Under the circumstances, the motor vehicles behavior has changed from homogeneous situations to heterogeneous conditions. The last traffic behavior type has reduced the vehicle speeds and the other modes, also made more congested (Zakaria et al. 2011). In further, travel time of the vehicles on urban roads under the traffic situation become worse. Regarding the heterogeneous traffic situation, some previous studies have been conducted. For examples, Chandra et al (2003) have studied the impact of lane width on roadway capacity in India. Minh et al. (2005) have founded that the speed distribution on urban roads in Hanoi follows the normal distribution. They also have grasped the speed characteristics of motorcycle such as speed, flow, and headway under the heterogeneous condition in the city. In comparing between the characteristics and the homogenous traffic condition in the city, they have founded that the empirical speed on different traffic composition and road characteristics exposed different speed level. In the other side, the average headway for all observed road locations has the same mean headway. Zakaria et al. (2011) have attempted to evaluate effects of interval time variations of the speed distribution in case on Makassar, Indonesia. In addition, Minh et al (2010) and Chandra et al (2003) have developed a motorcycle unit (MCU) as instead of passenger car unit (PCU) as representative unit of traffic for motorcycle-dominated traffic in Vietnam and India, respectively. Putranto et al (2011) have evaluated the performance of motorcycle lane in Jakarta and Sragen, Indonesia, where the exclusiveness of motorcycle lane did not significant effect to V/C ratio.in addition, impacts on the road environment of the heterogeneous traffic situation such as road traffic noise (Hustim& Fujimoto, 2012; 2013), vehicular emission (Aly, et al, 2016), and traffic accident (Halim, et al., 2017) also have been studied. Addressed to the travel time of vehicles on urban road networks, it is very useful information both for travelers and road authorities. In further, travel time modelling plays important rules in Advanced Traveler Information Systems (ATIS)(Mori, et al., 2014). It is the one important factor in traffic simulation in order to overcome traffic congestion problems based on traffic management system measures. Regarding this, in recently years, many studies on travel time estimation and prediction models have been developed. Most of them focused on travel times on freeways, not only model based approaches (Chen, et al, 2001; Chien, et al, 2002; Kwon, et al, 2005; Wei, et al, 2007), but also data-driven approaches (Wu, et al, 2004; editor@iaeme.com

3 A Travel Time Estimation Model of Private Cars in Urban Arterial Roads Based on Heterogeneous Traffic Innamaa, et al, 2005; Jie Yu, et al, 2008; Hinsbergen, et al, 2009). Those models have revealed satisfactory results for the traffic states along the route. However, the approach of the studies is less successful in applying on the urban road (Zheng, at al, 2010) due to the traffic behaviors on urban trips is significantly different than on freeways. Travel times of vehicles on urban roads at least are determined by four mechanisms (Zheng, at al, 2010): the driving speed on the urban roads; the queuing process before the intersection; the traffic control at intersections and parking movements; loading-unloading public transit at stops. These mechanisms lead to a consequency that the travel time (delay) is not characterized by a single value but by a certain travel time (delay) distribution (Zheng, at al, 2010). However, in many developing countries there is difficulty to measure and determine the vehicle travel time due to the equipment measure constraint such the loop detector costly etc. In this regard, the dynamic moving or driving cycle parameters of the vehicle moving on the road such as the average speed of the private car, the average speed of the car without idling mode, the acceleration, the deceleration, and the cruising time of the private carcould be utilized to describe the travel time phenomena of the vehicle, particularly for the private car type. In contributing on the travel time of vehicles research filed, especially on the heterogeneous traffic behavior in Indonesia, the present paper proposes and adopts the vehicle probe method to observe the characteristics of vehicle travel time under a heterogeneous traffic situation on the urban arterial roads in Makassar City, Indonesia. The rest of the present paper is organized as follows. Section 2 describes the study methods such the survey location, the equipment survey, and survey method of the travel time investigation. Section 3 presents the results of the travel time investigation for the light vehicle. The final section provides conclusions related to the results. 2. MATERIALS & EXPERIMENTAL PROCEDURES 2.1. The Urban Roads Location The data collection for the travel time measurement of the private cars in Makassar City, Indonesia was conducted in selected 26 urban roads which has status as urban arterial roads in the city. The study selected the urban roads in order to represent the various characteristics of the available arterial roads in the city. The characteristics of the selected arterial roads such as road types, road width, road shoulder width, and road length, are shown in Table The Equipment of the Travel Time Measurement The main equipment for the travel time measurement in this study consists of two equipment, i.e., a global position system (GPS) equipment, and a private car as a test vehicle. This study used GPS Garmin Etrex 30 to track the private car velocity in second by second along through the road for travel time survey based on probe vehicle. The test vehicle in this study used a passenger car with AVANSA type which produced by TOYOTA. We selected the test vehicle type as probe vehicle due to the vehicle type is dominant composition in the urban roads in Makassar City editor@iaeme.com

4 Sahrullah, Muh. Isran Ramli, Nur Ali and Ramli Rahim Table 1 Characteristics of the urban roads for the survey location No. Road names Road types Road width Road shoulder (m/lane) width (m) 1 Abd. Dg. Sirua 2/2 UD A.P. Pettarani 8/2 D 3.75 > Aroepala 4/2 D 3.00 > Bandang 4/2 D 3.00 < Boulevard 6/2 D 4.00 Shoulder < 0,5 m 6 Bulusaraung 3/1 UD 3.00 Shoulder < 0,5 m 7 Cendrawasih 4/2 UD 3.00 Shoulder 1,0 m 8 Dg. Tata 2/2 UD 6.00 Shoulder < 0,5 m 9 Sam Ratulangi 4/2 UD 3.00 Kerb> 2 m 10 WahidinS. 3/ Shoulder < 0,5 m 11 G. Bawakaraeng 4/1 D 3.00 Kerb< 0,5 m 12 Ahmad Yani 4/1 UD 3.00 Kerb< 0,5 m 13 Sudirman 4/2 UD 4.00 Kerb< 0,5 m 14 Hertasning 4/2 D 3.00 Shoulder < 0,5 m 15 UripSumohardjo 6/2 D 3.25 Kerb< 0,5 m 16 Malengkeri 2/2 UD 6.00 Shoulder 1 m 17 Masjid Raya 4/1 UD 3.00 Kerb< 0,5 m 18 Nusantara 4/2 D 3.00 Kerb< 0,5 m 19 Pengayoman 4/2 D 4.00 Shoulder < 0,5 m 20 Penghibur 2/1 UD 3.00 Shoulder < 0,5 m 21 Perintis K. 6/2 D 3.50 Shoulder > 2 m 22 Sulawesi 3/1 UD 3.00 Shoulder < 0,5 m 23 St.Alauddin 4/2 D 3.50 Kerb> 2 m 24 St.Hasanuddin 4/1 UD 3.00 Shoulder < 0,5 m 25 Veteran Selatan 4/2 D 3.50 Shoulder < 0,5 m 26 Veteran Utara 4/2 D 3.50 Shoulder < 0,5 m 2.3. The Measurement Method of the Private Car s Travel Time The travel time survey for the private cars based on the probe vehicle method which adopts a floating car survey method. The survey method used the vehicle test in order to capture the real-world traffic flow situation on the road. The probe vehicle approach is to determine the private car speed on the urban road network. The probe vehicle method is based on the collection of localization data, speed, directions of travel and time information from mobile source in the vehicle that are being driven. In this regard, the test vehicle and an active mobile source (such as GPS) act as the sensor for traffic flow on the roads. By applying the travel time measurement method using both equipment, GPS and the private car, we conducted the travel time survey for the private car on the twenty sixes urban roads in Makassar City. The travel time survey of the test-vehicle tracks from the starting point until the end point of each urban as the study location. The private car driver drives on the urban roads with the natural speed of the traffic flow. The driver drives the car at the ambient speed which the speed did not travel faster (overtaking more vehicles than overtook the test vehicle), or slower (being overtaken by more vehicles than were overtaken by the test vehicle) than the traffic flow speed. In the other hand, at the similar time, the other surveyor or an assistant which is riding the car, sets the GPS to record the private car speed second by second and the travel time over the road length. The survey method was repeated three times using the same private car for each traffic direction and for each traffic peak-hour period in capturing the variation of the road traffic editor@iaeme.com

5 A Travel Time Estimation Model of Private Cars in Urban Arterial Roads Based on Heterogeneous Traffic situation. The three peak-hour periods involve morning peak period, noon peak period, and evening peak period The Model Construction for the Private Car Travel Time The analysis data in order to construct the travel time model of the private car consists of some activity analysis. Firstly, the measurement data using the GPS were extracted from the GPS to a computer using a mapping software, then it was tabulated in the spreadsheet analysis. Secondly, the driving cycle parameters of the test vehicle were analyzed using the descriptive statistic analysis. There were five parameters of the driving cycle were analyzed, i.e., the average speed of the private car (V1), the average speed of the car without idling mode (V2), the acceleration (A), the deceleration (D), and the cruising time of the private car (C). Thirdly, the development of the model construction for the private car travel time. In this regard, this study used a multiple linear regression model approach. The study constrained to use only five parameters of the driving cycle as the independent variables in the model development. In this regard, the five parameters represent the dynamic motion of the private cars on the urban roads. The five parameters which taking account into the model involve the average speed of the private car (V1), the average speed of the car without idling mode (V2), the acceleration (A), the deceleration (D), the cruising time of the private car (C). The travel time model of the private car was constructed as the equation (1) below. Y = β 0 + β V1 X V1 + β V2 X V2 + β A X A + β D X D + β C X C (1) Where: Y is the travel time as the dependent variable; β 0 is a constant of the model; β V1, β V2, β A, β D, β C are the parameters of the model variables, i.e., the car average speed variable (XV1), the variable of the average speed without idling car (XV2), the car acceleration variable (XA), the car deceleration variable (XD), and the variable of the private car cruising time (XC), respectively. The present study utilized the Likelihood Maximization method in order to estimate the parameters values of the model. Regarding the data collection, this study calibrated three models of the private car travel time, i.e., morning peak period model, noon peak period model, and evening peak period model. The calibrated models were validated using observed travel time data on some urban roads in Makassar City. 3. RESULTS AND DISCUSSION 3.1. The Travel Time Condition of the Private Car on the Urban Roads Fig. 1 shows the survey results of the private car travel time in the twenty sixes urban roads in Makassar City. The travel time of the private car varied from 9.7 seconds until 51.0 seconds per-100 meters. However, majority of the urban roads have travel time for the private car around 15 seconds until 20 seconds per-100 meters. These travel times were determined from the dynamic motion of the private car in the urban roads. In this regard, the dynamic motion of the car involves the driving cycle parameters of the vehicle, i.e., the average speed of the private car (V 1 ), the average speed of the car without idling mode (V 2 ), the acceleration (A), the deceleration (D), and the cruising time of the private car (C) editor@iaeme.com

6 Sahrullah, Muh. Isran Ramli, Nur Ali and Ramli Rahim Figure 1 The travel time of the private car on the urban roads in Makassar City 3.2. The Dynamic Moving Parameters of the Private Car on the Urban Roads Table 2 shows the parameters values for the dynamic moving parameters of the private cars on the urban roads. Table 2 shows s that the average speed of the private car (V1), and the average speed of the car without idling mode (V2) varied from 12 Kmhr-1 until 29 Kmhr-1. However, the parameters values of the average speed of the private car is slightly bigger than the the average e speed of the car without idling mode. Furthermore, the acceleration (A), and the deceleration (D) values of the private car also fluctuated on interval ms-2. In addition, the sample of each road types for the probability density function (pdf), and the cumulative density function (cdf) of both parameters, the acceleration (A), and the deceleration (D) are showed in Figure 2. In further, the values of the cruising time of the private car (C) varied from 34 seconds until 208 seconds. Table 2 The Average values of the dynamic moving parameters of private cars Dynamic moving parametersof private cars The urban V roads/track 1 V 2 D A (Kmhr -1 ) (Kmhr -1 ) (ms -2 ) (ms -2 ) 8/2 D (A.P. Pettarani) Track Track Track /2 D (Boulevard) Track Track C (Sec) editor@iaeme.com

7 A Travel Time Estimation Model of Private Cars in Urban Arterial Roads Based on Heterogeneous Traffic Track /2 D (Pengayoman) Track Track Track /2 UD (Sudirman) Track Track Track /1 D (G. Bawakaraeng) Track Track Track /1 UD (Ahmad Yani) Track Track Track /1 UD (Sulawesi) Track Track Track /2 UD (Abd. Dg. Sirua) Track Track Track The Travel Time Model of the Private Car on the Urban Roads The calibration results of the travel time estimation model for three periods on a day, i.e., the morning period, the around noon period, and the afternoon period are showed in Table 3. In further, the validation of the three types of the travel time models is showed in Figure 2. Table 3 shows that the three estimation models have very good of the goodness of fit. These were indicated by the R 2 values of the three models which have 0.8 until 0.9. In further, the P-values of the model parameters indicated that the all variables taking account into the models were significant at various level for the morning period of travel time model. However, for the around noon period, and the afternoon period of the travel time models, the average speed of the private car (V 1 ), and the average speed of the car without idling mode (V 2 ), were not significant influence the travel time of the private car. In addition, mostly the signs of the parameters values followed the expected signs. In this regard, the signs of the parameters values for the average speed of the private car (V 1 ), and the average speed of the car without idling mode (V 2 ), across each other among the three models. Generally, the models have good indicators in order to represent the real-world phenomena of travel time for private cars in urban roads under heterogenous traffic situation editor@iaeme.com

8 Parameters symbols Sahrullah, Muh. Isran Ramli, Nur Ali and Ramli Rahim Table 3 Calibration results of the travel time models Estimated parameters of the travel time models Morning period Around noon period Afternoon period Parameter values P-Value Parameter values P-Value Parameter values P-Value β β V * β V * β D ** * * β A *** ** * β C * * * R N Note: the variables are significant at the level 95% (*), 90% (**), and 80% (***) Measured travel time (Seconds) RMSE : 1,109 Measured travel time (Seconds) RMSE : 2,188 Measured travel time (Seconds) RMSE : 1, Calculated travel time (Seconds) Calculated travel time (Seconds) Calculated travel time (Seconds) Figure 2 The validation results of the travel time models Figure 2 shows that the travel time estimation model for three periods on a day, i.e., the morning period, the around noon period, and the afternoon period, have good root mean square error (RMSE) values. The small values of the RMSE implicate that the empirical models of the travel time estimation have good validation. Regarding the calibration and validation results, the cruising time of the private car (C), the deceleration (D), andthe acceleration (A)have significantly influenced the travel estimation model for the private cars under heterogeneous traffic condition. In this regard, the cruising time of the private car (C) is more significant than both variables. In the other side, the deceleration (D) variable is more significant than the acceleration (A) variable. 4. CONCLUSIONS The travel time estimation model for private cars on urban roads under a heterogeneous traffic situation has been developed using an empirical model approach, i.e., the multiple linear regression. The travel time estimation model was constructed from dynamic moving parameters of the private cars on the urban roads such as the average speed of the private car, the average speed of the car without idling mode, the acceleration, the deceleration, and the cruising time of the private car. The variables values were measured on the arterial urban roads which have the heterogeneous traffic situation, in Makassar City, Indonesia, as a case study. The measurement applied the probe vehicle survey approach, i.e. the floating car method, using a GPS and a test private car. The cruising time of the private car, the deceleration, andthe acceleration have important rule in the travel estimation model for the private cars under heterogeneous traffic condition editor@iaeme.com

9 A Travel Time Estimation Model of Private Cars in Urban Arterial Roads Based on Heterogeneous Traffic However, the three variables have various significant level in influencing the travel time estimation. Briefly, we expect that the results provide an empirical model in estimating travel time of the private car on urban roads for heterogeneous traffic condition. REFERENCES [1] Aly, S.H., Selintung, M., Wunas, S., Ramli, M.I Macroscopic Analysis of Heterogeneous Traffic Behavior on Divided Urban Roadway. Proceeding of the 14th FSTPT International Symposium. [2] Aly, S.H., Ramli, M.I, and Sumi, T Driving Cycle of Passenger Cars on Heterogeneous Traffic Situations: Case Study on an Urban Road in Makassar, Indonesia. Proceeding of the 8th International Symposium on Lowland Technology. [3] Chen, M., S. I. J., Chien. Dynamic freeway travel time prediction using probe vehicle data: Link-based vs. Path-based. Transportation Research Record (1768)13, [4] Chien, S. I. J., and C. M., Kuchipudi. Dynamic Travel Time Prediction with Real-time and Historical Data. 81st Annual meeting of Transportation Research Board, Washington D.C, [5] Chandra, S., and Kumar, U Effect of Lane Width on Capacity under Mixed Traffic Condition in India. Journal of Transportation Engineering, Mar. Apr., [6] Fangfang Zheng, and Henk van Zuylen, (2010), Comparison of Urban Link Travel Time Estimation Models Based on Probe Vehicle Data. Traffic and Transportation Studies, pp [7] Hasmar Halim, Sakti AdjiAdisasmita, Muh. Isran Ramli and Sumarni Hamid Aly, (2017), The Pattern of Severity of Traffic Accidents on Traffic Conditions Heterogeneous. International Journal of Civil Engineering and Technology, 8(4), pp [8] Hustim, M., Anai, K., and Fujimoto, K Survey on Road Traffic Noise in Makassar City, Indonesia. Proceeding of 40 th International Congress and Exposition on Noise Control Engineering 2011, INTER-NOISE [9] Hustim, M. and Fujimoto, K Road Traffic Noise under Heterogeneous Traffic Condition in Makassar City, Indonesia. Journal of Habitat Engineering and Design, Vol. 4, No. 1, pp [10] Hustim, M. and Fujimoto, K Road traffic noise reduction using TDM-TMS strategies in Makassar city, Indonesia. Journal of Environmental Engineering (Transaction of AIJ), 78(689), pp [11] Innamaa, S. Short-term prediction of travel time using neural networks on an interurban highway. Transportation, 32: , [12] Jie Yu, G. L. C., H. W. Ho, and Yue Liu. Variation Based Online Travel Time Prediction Using Clustered Neural Networks. 11th International IEEE Conference on Intelligent Transportation Systems, Beijing, [13] Kwon, J., K., Petty. A Travel Time Prediction Algorithm Scalable to Freeway Networks with Many Nodes with Arbitrary Travel Routes. TRB 84th Annual Meeting, Washington, D.C., [14] Minh, C.C., Sano, K., and Matsumoto, S The Speed, Flow and Headway Analyses of Motorcycle Traffic. Journal of the Eastern Asia Society for Transportation Studies, Vol.6, pp editor@iaeme.com

10 Sahrullah, Muh. Isran Ramli, Nur Ali and Ramli Rahim [15] Minh, C.C., Sano, K., Mai, T.T., and Matsumoto, S Development of Motorcycle Unit for Motorcycle-Dominated Traffic. Journal of the Eastern Asia Society for Transportation Studies, Vol.8, pp [16] Putranto, L.S., Suardika, G.P., Sunggiardi, R., Munandar, A.S., and Lutfi, I The Performance of Motorcycle Lanes in Jakarta and Sragen. Proceeding of the 9th Conference of the Eastern Asia Society for Transportation Studies. [17] Sumarni Hamid Aly and Muhammad Isran Ramli, (2016). A Development of MARNI 12.2 Model: A Calculation Tool of Vehicular Emission for Heterogeneous Traffic Conditions. Journal of Engineering and Applied Sciences, 11: [18] Wei, C. H. &Y. Lee. Development of freeway travel time forecasting models by integrating different sources of traffic data. IEEE Transactions on Vehicle Technology, 56 (6) , [19] Usue Mori, Alexander Mendiburu, MaiteÁlvarez& Jose A. Lozano (2014): A review of travel time estimation and forecasting for Advanced Traveller Information Systems, Transportmetrica A: Transport Science, DOI: / [20] Van Hinsbergen, C. P. IJ., J. W. C. Van Lint, H.J. Van Zuylen. Bayesian training and committees of State Space Neural Networks for online travel time prediction. Transportation Research Record: Journal of the Transportation Research Board, 2105: , [21] Wu, C., C. W. Chia, D.S.SU, M.H. Chang, and J. M. Ho. Travel Time Prediction with Support Vector Regression. IEEE Transactions on intelligent transportation systems, 5 (4)6, [22] Zakaria, A., Aly, S.H., Ramli, M.I Distribution Model of Motorcycle Speed on Divided Roadway in Makassar. Proceeding of the 14th FSTPT International Symposium. [23] Akshay Jadhav, Deepak Anchule, Shekhar Bade and Prof. Anuradha Pansare, Optimized Solutions for Resolving Traffic congestion At University Circle, International Journal of Civil Engineering and Technology, 7(2), 2016, pp [24] Samar Patni, V. S. Landge and Sanket Gupta, Motor Vehicle Traffic Congestion Costing in Nagpur City, International Journal of Civil Engineering and Technology, 8(4), 2017, pp editor@iaeme.com