Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007

THE IMPACT OF PUBLIC TRANSPORT ARRIVAL RATE AND STOP TIME IN MODELLING AND ANALYZING A SIGNALIZED INTERSECTION BY USING MICRO SUMULATION AND ANALYTICAL SOFTWARE

Gusri YALDI Wen Long YUE Lecturer Senior Lecturer, Program Director Civil Engineering Department Transport Systems Centre Polytechnic of Andalas University University of South Australia Kampus Politeknik Unand North Terrace, Adelaide Limau Manis, South Australia 25161 SA 5001 Fax: +62-751-72576 Fax: +61-8-83021880 E-mail: [email protected] E-mail; [email protected]

Elvi Roza SYOFYAN Lecturer Civil Engineering Department Polytechnic of Andalas University Kampus Politeknik Unand Limau Manis, Padang 25161 Indonesia Fax: +62-751-72576 E-mail: [email protected]

Abstract: The application of micro simulation traffic modeling and analyzing software has been widely spread in many countries, including the developing countries. Before that, many developing countries have been used micro analytical software, for example aaSIDRA. It seems it is more difficult to use micro simulation software in developing countries due to there are more complex problems compared to developed countries, as the models were developed based on the travel behavior of the country. A study to model and analyze a signalized intersection has been undertaken in Padang, West , Indonesia. It used CUBE Dynasim, a relatively new micro simulation software, and aaSIDRA. Throught the application of the two models, it has been found that CUBE Dynasim tends to generate lower approach flows compared to the real data while aaSIDRA generates higher lane capacity than the demand. Public transport arrival rate and stop time were reduced with by 10 percent incrementally in order to investigate this case. However, CUBE Dynasim still generates traffic flow which is below the real data.

Key Words: Traffic flow, Public transport arrival rate, Public transport stop time

1. INTRODUCTION

The majority of developing countries, including Indonesia, are likely having the same problems related to the urban transport systems. The travel demand increases as the urbanization rate increases which could reach as twice quicker as in the United States (Morichi, 2005). It is even higher compared to the countries in Europe. Due to the increases, it has caused difficulties in modeling and analyzing their transport systems in developing countries compared to in developed countries where the system has been well-established, Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007 particularly when the modeling and analysis is undertaken by using micro simulation instead of micro analytical traffic analysis software. Hence, the application of micro simulation traffic analysis and modeling software which have been successfully in developed countries could generate different results when used in developing countries.

The use of microscopic traffic simulation plays major roles in the analysis and evaluation of transport systems. It is due to its ability to analyze a transport system based on each vehicle properties and operations. It can analyze the interaction between vehicles in the systems and between vehicles and the infrastructure as well. Therefore, microscopic traffic simulators can be considered as a suitable tool in analyzing various transport operations (Barcelo et al. 2003). It can reproduce a significant level of accuracy and capture the interactive impacts among transport elements in a system. Besides, it can generate outputs which show the variations of particular transport system parameters. For example, it can produce the average travel speed and its variations so that the profile of the speed within certain period of time can be evaluated. Hence, through a simulation analysis the practitioners are able to estimate the likely outcomes in the system after some alternative changes are applied. Therefore, the best scheme among the proposed planes can be selected accordingly (Gomes et al. 2003).

However, the capability of a particular traffic simulator to accommodate traffic planners’/engineers’ needs in analyzing a transport system must be equipped with its ability to generate remarkable, accurate and precise outputs. More precisely, the software must have a tool that can be used to adjust the factors to calibrate the analysis based on local conditions. Dowling et al. (2004) explained that calibration process is important since the appropriate analysis parameters can be selected according to the local traffic operation conditions. Some of data which is required to be calibrated are, for examples, traffic volumes, average travel speeds, peak time factor, and acceptable gap.

The main aim of this paper is to discuss the use of CUBE Dynasim and aaSIDRA in modeling and analyzing a signalized intersection located in the Central Business District (CBD) of Padang, -Indonesia. Further, it focuses on the impact of Public Transport (PT) arrival rate and stop time to the traffic flow. The comparisons regarding the results generated by both software are then compared in order to see the performance differences of the two software in analyzing and modeling a signalized intersection. CUBE Dynasim is an event- based software with stochastic and dynamics outputs. It has tools which can be used to model the real transport system including the application of Intelligent Transport System (ITS) facilities, for instance, actuated traffic signals. It is multimodal traffic simulator software and able to import file or data from CAD, GIS and other databases as well as other traffic analysis programs. This micro simulation software also has some tools to calibrate or adjust the traffic parameters based on the local traffic data.

On the other hand, aaSIDRA is micro analytical traffic software and has been used in more than 80 countries, predominantly in the USA and Australia. It has the ability to analyze an intersection with up to 8 legs with options of two-way road, one-way approach or one-way exit. In addition, it can calibrate the analysis based on local conditions and compatible with the Highway Capacity Manual (HCM), for example the signal as well as the road conditions. Moreover, it can determine the optimum cycle time which is unable for micro simulation software to perform. aaSIDRA can count the impact of on-street parking and bus stopping on the system, however, it is only for on-street parking and bus stopping which is located on the approach lane. Therefore, this could be a weaknesses of aaSIDRA compared to CUBE Dynasim. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007

This study will involve only one signalized intersection in Padang CBD, West Sumatra, as a test case. It is a-four leg intersection and controlled with pre-timed signal. It is an intersection between St and St as shown in figure 1. There are only three approaches namely Sudirman St, Agus Salim St and Bagindo Aziz Chan St on the East, South and West approach respectively. The North approach, in this case Pasar St, is exit only and an one-way road. The data are collected by undertaking traffic surveys including traffic volumes and distributions, signal timings, bus headways and on-street parking. Then, a comparison has been made addressing the ability of each software in modeling the intersection as well as the output generated by the two software. The output comparisons will focus on the traffic flow with various of PT arrival rates and stop times.

Some limitations are applied in the research. Those are, for example, the vulnerable road users, including non-motorized transport modes were ignored. The adjacent minor junctions were also ignored. The transport modes are categorized into three groups, namely Passenger Car (PC), Heavy Vehicle (HV) and Motorcycles (MC) based on the local code, in this case Manual Kapasitas Jalan Indonesia (MJKI). All motorcycles were converted into passenger car unit by multiplying them with 0.2 while PC-size public transports were converted into heavy vehicles by multiplying them with 1.3. The conversions were made based on the local code (MKJI) and were undertaken before entering the data into both software.

Figure 1 Intersection layout

2. COMPARISON OF CUBE Dynasim AND aaSIDRA OUTPUTS

The aim of micro simulation modeling and analysis is, for example, to evaluate the traffic movement during a definite period of time. The impact of every intervention such as public transport stopping, traffic signals and crossing pedestrians to the traffic flow can be investigated. Firstly, both UBE Dynasim and aaSIDRA models were developed. All required data used in both models are the same, including the traffic volume and distribution, intersection geometry and phase and signal timings. There are three bus stops in the model. One bus stop located on Bagindo Aziz Chan approach. There is also one bus stop on Bagindo Aziz Chan exit. Another one is on the Sudirman approach. There is no bus stop located on Agus Salim approach. The results of the modeling and analysis are discussed below.

In CUBE Dynasim, the traffic flow is considered as instant flow. The instant flow for each lane from a certain origin and destination can be collected by setting up the data collector on the simulation objects. After running the model, the instant flow for each lane is generated by CUBE Dynasim. The proximity of the simulation outputs to the real data is determined by the Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007 number of simulation undertaken in the analysis. In this study, 10 runs are adopted to form a statistical summary. CUBE Dynasim produces the simulation results in two kinds of statistical outputs which are tables and graphs. Figures 2 and 3 are the examples of graphs resulted from CUBE Dynasim. Those graphs portray the approach, exiting and lane flow at the observed intersection within one hour for each direction. Like other micro simulation traffic analysis software, CUBE Dynasim can generate the output as small as at one second interval. The output in this research is set to be reported for an interval of 900 seconds or 15-minute interval. The simulation time is 60 minutes.

The highest 15-minute flow on each approach tends to vary. For instance, Sudirman approach had the highest at 8.00 am then declined until 8.45 am. Afterward, it started to increase. At the mean time, Agus Salim and Bagindo Aziz Chan approaches had a contrary trend compared to each other. In general, however, all approaches had slightly a different trend to the real data. The possible cause is, for instance, the local peak hour-period. From one-week traffic survey, it has been found that the peak time is likely between 08.00-08.30. It could be different from the local conditions of where the software was developed. Another factor is the simulation itself as two fundamental concepts in the simulation modeling are the analysis is stochastic and dynamics.

Figure 2 CUBE Dynasim approach flow

The lane flow of each approach generated by CUBE Dynasim is reported in table 1 while figure 3 illustrates the exiting flow on each approach. It can be seen in table 1 that CUBE Dynasim generates a lower number of vehicles than the real data. Bagindo Aziz Chan approach, for example, has only 737 vehicles per hour or 14 percent lower than reality. In contras, aaSIDRA calculates that the degree of saturation (X) of each lane on Bagindo Aziz Chan approach is much lower than one, or, the capacity is much higher than the demand. Therefore, aaSDIRA could have resulted outputs close to the real conditions of the intersection. Furthermore, there were more motorists traveled to the western area than others as shown by figure 3. The lowest was to the southern area. This time series-volume illustrates a different figure compared to the real data (see figure 4). The figure demonstrates that the highest exiting flow is on the Bagindo Aziz Chan approach or east approach. However, the real data has the highest exiting flow on the Sudirman approach or west approach.

The differences between CUBE Dynasim to the real data and aaSIDRA output is likely triggered by several possible causes; those are the high number of PT arrival with relatively high average stop time, the number and type of lane discipline and the attitudes of motorist on the road. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007

As micro analytical traffic analysis software, aaSIDRA takes into account the impact of public transport stop frequencies. However, it is only on the approach lanes. Besides, it is not explicitly considering the stop time of public transport. From aaSIDRA analysis, it was found that the degree of saturation of all lanes at the intersection is less than 1 (see figure 5). It means that the intersection still has spare capacity, or, the traffic volume is lower than the capacity of the intersection. Based on the traffic observation at the analyzed intersection, the traffic flow was moving fairly stable. Therefore, it can be assumed that aaSIDRA could have generated a similar result as the real condition of the intersection.

Figure 3 CUBE Dynasim exiting flow

A1 A4 (Pasar) (Sudirman) 22% 33%

A3 (Bagindo A C) A2 (Agus 31% Salim) 14%

Figure 4 Real data exiting flow

Table 1 CUBE Dynasim lane flow, field data and the differences Traffic flow (vec/h) Lane category Difference Cube Dynasim Field data Sudirman approach LT 367 363 1% TH1 337 352 -4% TH2 327 352 -7% RT 94 93 1% Total 1130 1160 -3% Agus Salim approach LT 216 219 -1% TH 79 81 -2% RT 576 642 -10% Total 874 942 -8% Bagindo Aziz Chan approach LT 380 465 -18% TH 300 335 -10% RT 65 66 -2% Total 737 866 -14% Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007

Figure 5 aaSIDRA intersection’s degree of saturation

Figure 6 PT queuing on left lane of Bagindo Aziz Chan approach

CUBE Dynasim is a micro simulation traffic modeling and analysis software. It generates traffic movement at the intersection by generating two and three dimension of animations. From this simulation, the impact of public transport arrivals and stop time can be observed clearly. For example, there is a long queue of public transports on the most left lane of Bagindo Aziz Chan approach as illustrated by figure 6. The long queue is formed since the public transport stop time is higher as well as its arrival rate. It could reach up to 405 seconds or equals to about 6.5 minutes for the stop time, while the arrival rate is 123.

This could block the movement of other road users, including the other public transport vehicles behind and hence the traffic flow, particularly left turn and through vehicles, decreased. Besides, the available lanes on this approach are shared lanes. Therefore, some vehicles with destinations of Pasar and Sudirman approaches are blocked. This situation could occur when there is a high number of public transport arrival and high stop time. Nonetheless, the vehicles behind the queuing public transport should change the lanes quickly and proceed to their destinations as normally happened in the real situation.

A study conducted by Yaldi and Yue (2006) has demonstrated that both aaSIDRA and CUBE Dynasim generated fairly the same results to the real data investigated. That study was conducted in Adelaide, South Australia. In that study, the maximum stop time of PT was 44 seconds or less than one minute. In addition, there were only 8 PT arrivals and majority of lanes were not shared lanes. Therefore, it can be assumed that, high number of arrivals and maximum stop time could drop off the intersection capacity of CUBE Dynasim model.

To investigate this case, both CUBE Dynasim and aaSIDRA base models were modified. Firstly, only CUBE Dynasim model was modified. The number of PT arrival remains the same, yet, the stop time was incrementally reduced. The increment is 10 percent. The purpose is simply to seek the impact of various stop time to the approach and exiting flow of Bagindo Aziz Chan and Sudirman approach. The aaSIDRA model was not modified, since aaSIDRA only considers the frequency of bus stops. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007

Figures 7, 8 and 9 show the traffic flows on Sudirman, Agus Salim and Bagindo Aziz Chan approaches respectively. It can be seen that the traffic flow slightly fluctuates; however, the variation is relatively small, it increases about one percent. Hence, the length of stop time could affect the traffic flow in CUBE Dynasim model; however, it is not significant.

Another attempt to investigate this case is by reducing the PT arrival rates on all approaches. The PT arrival rate is set to be 30, 40, 50, 60, 70, 80, 90 and 100 percent of the real data. The modification is applied in both aaSIDRA and CUBE Dynasim models. The PT stop time remains the same as the original data. The purpose is to investigate the proximity of CUBE Dynasim results to the real data and aaSIDRA results. Furthermore, the investigation will highlight the Bagindo Aziz Chan and Sudirman approaches only, particularly the left and through lanes of Bagindo Aziz Chan approach and through lane of Sudirman approach. Special treatment will be undertaken on the Agus Salim approach. It is due to the different between CUBE Dynasim flow and the real data are predominantly on those lanes. The results are depicted by figure 10, 11 and 12.

800 700 600 500 LT 400 TH 300 RT 200

Taffic flow (veh/h) 100 0 050100150 Percentage of PT Stop Tim e (%)

Figure 7 Traffic flow on Sudirman approach with PT stop time reduction

700 600 500 LT 400 TH 300 RT 200 100 Traffic flow (Veh/h) flow Traffic 0 0 20 40 60 80 100 120 Percentage of PT Stop tim e (%)

Figure 8 Traffic flow on Agus Salim approach with PT stop time reduction

Figures 11 and 12 illustrate that CUBE Dynasim tends to generate almost the same lane flow on Bagindo Aziz Chan approach than before even though the PT arrival is reduced. The line on the graph tends to be flat and to be intersected with the real data when the PT arrival reduction is 100 percent, or, there is no PT arrival in the model. Meanwhile, aaSIDRA generated the lane capacity which increases sharply when the PT arrival is reduced. It occurs on both left and through lanes of Bagindo Aziz Chan approach as well as on through lane of Sudirman approach (see figures 10, 11 and 12). Thus, it can be assumed that, the higher the PT arrival rate, the bigger the deviation of the traffic flow generated by CUBE Dynasim to the Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007 real data and aaSIDRA outputs. CUBE Dynasim model would generate the same lane flow to the real data when the PT arrival rate is zero.

500

400 LT 300 TH 200 RT 100 Traffic Flow (Veh/h) Flow Traffic 0 0 50 100 150 Percentage of PT Stop Tim e (%)

Figure 9 Traffic flow on Bagindo Aziz Chan approach with PT stop time reduction

1200

1000

800 CUBE Dynasim 600 Field data aaSIDRA 400 Traffic flow(vec/h) 200 0 0 50 100 150 Percentage of PT Arrival reduction (%)

Figure 10 Bagindo Aziz Chan left turn flow with various PT arrival rate reductions

900 800 700 600 CUBE Dynasim 500 Field Data 400 aaSIDRA 300

Lane flow (vec/h) flow Lane 200 100 0 0 20406080 Percentage of PT Arrival reduction (%)

Figure 11 Bagindo Aziz Chan through lane flow with various PT arrival rate reductions

The last attempt is by modifying the CUBE Dynasim once more. As discussed above, the CUBE Dynasim was modified on the basis of stop time and PT arrival by 10 percent incrementally. There were no significant changes related to the traffic flow on all approaches. Hence, it could be assumed that PT stop time and arrival could reduce the approach flow significantly. CUBE Dynasim generated the same approach flow as real data when PT was removed from the model. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007

1800 1600 1400 1200 CUBE Dynasim 1000 Field Data 800 aaSIDRA 600

(vec/h) flow Lane 400 200 0 0 50 100 150 Percentage of PT arrival reduction (%)

Figure 12 Sudirman through lane flow with various PT arrival rate reductions

However, the Sudirman approach flow tends to increase until the PT stop time equals to 50 percent of the original data. It was 1126 vehicles per hour. In the mean time, CUBE Dynasim also generated 1126 vehicles per hour on the Sudirman approach when the PT arrival was reduced 30 percent and it is the same as the real data. It may relate to the type and number of lane available on this approach. The Left and right lanes are shared lanes. In addition, the right lane is also a short lane. The competition in using road space is higher. Each road users would compete in a limited space and time to reach their destination. Therefore, it is possible for CUBE Dynasim to generate the same Sudirman approach flow as the real data when the PT stop time is reduced by 50 percent and there is only 30 percent of PT arrival in the model as there would be lower number of vehicles in the network. Besides, the PT would stop to serve the passengers is a shorter time and hence it could cross the intersection quicker and reduce the blockages. However, it remained constant on other approach. Even though there was an increasing flow on the Bagindo Aziz Chan approach, yet, it was still below the real data. The increasing was ended when the PT stop time reduction reached 50 percent.

Based on these findings, the CUBE Dynasim is modified at a third time. The PT stop time is taken as 50 percent of the original data while the PT arrival is reduced by 10 percent incrementally. The purpose is to investigate the traffic flow generated by CUBE Dynasim when the PT stop equals to 50 percent of the real data. It also sees what would happen to the traffic flow at the intersection, particularly on Sudirman approach when the PT stop time is only 50 percent of the real data with a various PT arrival rates. It is expected that shorter stop time could enhance the capacity. Further, when would the approach flow reach the same level as the real data could be found? The results are given on figures 13, 14 and 15.

1170 1160 1150 Real data 1140 CUBE Dynasim 1130 Intersecting point 1120 1110 Approacg flow (vec/h) flow Approacg 1100 0 20406080100 Percentage of PT stop time reduction (%)

Figure 13 Sudirman approach flow with 50% reduction of PT arrival rate Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007

880 860 840 820 Field data 800 CUBE Dynasim 780 760 Intersecting point 740 720 Approach flow (veh/h) flow Approach 700 0 20406080100 Percentage of PT stop time reduction (%)

Figure 14 Bagindo Aziz Chan approach flow with 50% reduction of PT arrival rate

960 940 920 Real data 900 CUBE Dynasim 880 860 Approach flow (vec/h) flow Approach 840 0 20406080100 Percentage of PT stop time reduction (%)

Figure 15 Agus Salim approach flow with 50% reduction of PT arrival rate

It can be seen that the traffic flow on all approach are fluctuating. There is an increasing flow on the Sudirman approach. Nevertheless, it tends to decrease when the PT arrival is about 55 percent of the real data as shown by figure 13. On the other hand, the approach flow on the other approaches remains the same as before as illustrated by figures 14 and 15.

Based on the first, second and the third modifications, it can be presumed that the CUBE Dynasim could not generate the approach flow at the same number of the real data due to several factors. Those are likely the high stop time and PT arrival rates. Besides, the majority of lanes are shared lanes. As micro simulation software, CUBE Dynasim takes into account all events occurred during the simulation. For example, the motorists following behind a public transport would change the lane quickly. Than, it would overtake that public transport when they perceive the proceeding public transport would stop as happened in the real situation.

However, the acceleration and the reaction time factors embedded in the CUBE Dynasim software seems lower compared to the real local motorists’ aggressiveness. As discussed above, CUBE Dynasim would generate the same traffic flow as the real data when the PT arrival rate is zero. It began to generate lower traffic flow when the stop time reaches 50 percent and the public transport arrival rate equals to 30 percent of the original data. In other word, the aggressiveness of the motorists in CUBE Dynasim model is likely lower than in real life, particularly when it compared with the local motorists’ behaviors. Thus, it generated lesser flow when there were more interruptions in the system. With those percentages, the motorists still could travel at the intersection which is the same as the real situation. Yet, when the stop time is higher than 50 percent and the PT arrival rate is more than 30 percent of Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007 the original data, CUBE Dynasim produced lesser traffic flow compared to the real one. Therefore, the differences could be caused by some simulation factors such as the reaction time, acceleration and deceleration and driver aggressiveness factors built in the CUBE Dynasim software.

At the mean time, both CUBE Dynasim and aaSIDRA demonstrate that the number of lanes on each approach could reduce the intersection capacity as occurred on Agus Salim approach. There are only two lanes on the Agus Salim approach. The left lane is available for left, through and right turn movement, while the right lane is only for right turners. From traffic survey, it was found that the highest movement was right turns which are 642 vehicles per hour (see table 1). Even though there is no bus stop available on this approach, the number of PT using this approach is quite high. There were 61 PTs traveling on this approach with a destination of Sudirman approach.

Since there is only one lane exclusively for right turn movement, the only one left lane was also used by all transport modes went to the eastern area and must share with though and left turn vehicles. It seems that the available lanes could not provide sufficient space for right turners. Table 1 show that CUBE Dynasim generates only 576 motorists which could complete their movements or 10 percent lower than the real data. There is a long queue on that approach which could not move to their destination as shown by figure 16.

It has been discussed above that the PT arrival reduction did not increase the traffic flow on Agus Salim approach. It is still much lower compared to the real data. Meanwhile, the lowest degree of saturation generated by aaSIDRA for the same approach is 0.945 as depicted by figure 5. It means the capacity of the lane is still higher than the traffic flow, however, it is close. Both models could generate theoretically true capacity of that approach based on the real traffic, intersection and phasing and signal timing conditions. Therefore, the difference, particularly the CUBE Dynasim outputs, to the real data is not likely caused by PT arrival and its stop time. Yet, it could be originated by the attitudes of the motorist.

Form site observations at the intersection, it has been found that some motorists were still crossing the intersection even though the signal was red. Some also used the exit lane. In addition, the vehicles were queuing not only inside of each lane, but also between the lanes since the lane widths are 3.3 and 4.3 meters for left and right lanes respectively. To certain extend, this could increase the capacity as happened in the real data. However, it could direct the motorist to poor behavior of lane discipline since one of the lane width is more than 3.7 meters (Taylor et al, 2000). Then, it could caused accidents which may lead to a sever congestion.

To deal with this case, an additional lane on that approach would increase the approach capacity and enhance the road users’ attitudes. aaSIDRA model is modified to find the solution. The lane width of available lane is reduced. It becomes 2.6 and 2.5 meters for left and right lanes. There is still 2.5 meter left and is used as a new lane in the middle of the exiting lane. Therefore, there are three lanes available on Agus Salim approach. The new proposed intersection geometry can be seen on figure 17.

The modification results in lower degree of saturation, particularly on the Agus Salim approach. The maximum degree of saturation on the approach is 0.731 or 23 percent of that before. It means there is more space for vehicles to cross the intersections and additional lane has increased the approach capacity. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007

Figure 16 Left turn vehicles blocked by other road users

Figure 17 New proposed intersection layout

Figure 18 Degree of saturation of new intersection layout Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007

3. CONCLUSIONS AND RECOMMENDATIONS

3.1 Conclusions

There are conclusions which can be drawn from the study. Some lane flows generated by CUBE Dynasim are lower than the real data, particularly on left lane of Bagindo Aziz Chan approach and on the right lane of Agus Salim approach. The differences are 18 and 10 percent respectively. It thus affects the approach flow. Bagindo Aziz Chan approach has about 14 percent below the real data while it is 8 percent on Agus Salim approach. On the other hand, aaSIDRA generated results which are close to the real data. The degrees of saturation on all lanes are below 1, yet, it is 0.945 on the Agus Salim approach. It was presumed that CUBE Dynasim generated lower approach flow compared to the real data due to the high number of public transport (PT) arrival as well as the length of stop time to serve the passenger which is high.

To deal with this situation, the CUBE Dynasim and aaSIDRA models were modified. The PT arrival rate and stop times were incrementally reduced by 10 percent. However, CUBE Dynasim were still generating approach flow mostly the same as before the modification undertaken. It is happened on all approaches. In contras, aaSIDRA generated lower degree of saturation when the PT arrival rate is reduced. Based on these findings, it can be conclude that the differences between CUBE Dynasim results to the real data and aaSIDRA are likely related to the PT arrival rate and stop time.

However, CUBE Dynasim would generate an increasing traffic flow when PT stop time is reduced by 50 percent of the original data. Further, it would generate the same traffic flow as the real data when the PT arrival rate is equal to 30 percent of the real data. It is, perhaps, due to the number of lanes available on each approach and the simulation factors embedded in the CUBE Dynasim software. There is likely a lower drivers’ reaction time and aggressiveness model embedded in the CUBE Dynasim compared to the real local motorists’ behaviors. As micro analytical traffic analysis and modeling software, aaSIDRA could generate the same outputs as the real intersection conditions since it does not consider drivers’ reaction time, acceleration and deceleration and aggressiveness.

3.2 Recommendations

Some findings were obtained from this study, for example, the impact of PT arrival and stop time to the traffic flow. Furthermore, the number of lanes seems significantly reduced the capacity of Agus Salim approach. When the number of available lane is increased, the degree of saturation of each lane becomes much lower than before. There is a possibility that CUBE Dynasim would generate the same traffic flow as the data when there are additional lanes in all approaches. It based on a study conducted by Yaldi and Yue (2006). Most lanes in that study were not a shared lane.

REFERENCES

Barcelo, P., Codina, E., Casas, J., Ferrer, J. L. & Garcia, D. (2003) Microscopic traffic simulation: A tool for the design, analysis and evaluation of intelligent transport systems, TSS, Paris. Directorate General Bina Marga. (1997), Manual Kapasitas Jalan Indonesia, . Proceedings of the Eastern Asia Society for Transportation Studies, Vol.6, 2007

Dowling, R., Skabardoms, A. & Alexiadis, V. (2004) Guidelines for applying microsimulation software. Technical Report No. FHWA-HRT-04-040, Washington D.C. Gomes, G., May, A. & Horowitz, R. (2003) A micro simulation model of a Congested freeway using VISSIM, University of California, Berkeley. Morichi, S. (2005) Long-term strategy for transport system in asian megacities, Journal of the Eastern Asia Society for Transportation Studies, Vol. 6, 1 – 22. Taylor, M. A. P., Bonsall, P. W. & Young, W. (2000) Understanding Traffic Systems: Data, Analysis and Presentation, 2nd edition, Ashgate Publishing Company, Aldershot, England. Yaldi, G. (2006) The use of CUBE Dynasim and aaSIDRA in modeling and analyzing a signalized intersection, Minor Thesis, School of Natural and Built Environment, University of South Australia, Adelaide. Yaldi, G. & Syofyan, E. R. (2007) Investigating local traffic parameters to analyze a signalized intersection by using CUBE Dynasim and aaSIDRA. Research Report No. 282/N.04.1-PG/2006, TPSDP Batch III Project, Polytechnic of Andalas University, Padang, Indonesia. Yaldi, G. & Yue, W. L. (2006) The use of CUBE Dynasim and aaSIDRA in analyzing a signalized intersection, ATRF06 proceeding paper, Queensland, Australia.