Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

Evaluation of and Signalized Designs using Microsimulation

Minh Le KIEU Clas RYDERGREN Master Student Assistant Professor Linkoping University Linkoping University Institute of Technology Department of Science and Technology SE-601 74 Norrkoping SE-601 74 Norrkoping Sweden Sweden E-mail: [email protected] E-mail: [email protected]

Abstract: With increasing urban flows and increased congestion, the question of design modifications and intersection reconstruction is a common problem for urban traffic managers. In this paper we compare and evaluate the design of a roundabout and intersection in the Norrkoping city, Sweden. The intersection has been simulated with the traffic microsimulation program Aimsun, with observed data from afternoon rush-hours traffic. A roundabout, a signalized intersection design and a “bowtie” intersection design have been proposed and evaluated using results from the simulation. The results from the simulation study indicate that a roundabout design performs better in terms of the travel time compared to a traditional signalized intersection design. A hybrid version of the roundabout and the signalized intersection, a bowtie intersection, shows better performance in terms of travel time passing the junction than both the roundabout and the signalized intersection.

Key Words: Roundabout, Signalized intersection, Microsimulation, Sweden

1. BACKGROUND

Traffic jams due to congestion in intersections seems to be an almost chronic problem in many metropolitan areas. This stresses the importance of choosing a good design for intersections, either at the time of construction, or when making modifications. When the funding for reconstruction work does not allow for grade-separated designs, the two dominant choices are a roundabout or a signalized intersection.

Traffic managers and scientist are discussing on which is the better intersection design in central city area. According to Younes (2000), signalized intersections have better performance in increased traffic demands and better traffic control in high traffic flow urban . The Engineering Division of Kansas, U.S estimated that a roundabout project would cost $735,855 in constructing compared to $707,492 for a signalized intersection (Alisoglu, 2010). However, some others studies believe that offer continuous and faster traffic flow and less traffic jam than signalized intersection (NCHRP Report 572, 2007) The official guideline FHWA (2000) from the U.S Federal Administration mentions that roundabouts are also believed to be safer for vehicles (Younes, 2000). Shoki et al. (2010) used the software aaSIDRA to compare the design of roundabout and signalized intersection in two intersections in Selangor, Malaysia and showed that the level of service had increased from F to B and C (Highway Capacity Manual, 2000) by replacing a signalized intersection by a roundabout. Especially in intersections where the left turning proportions are relatively high, roundabouts can handle left turn traffic better than the signalized intersection (FHWA, 2000). Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

The purpose of this paper is to present results from intersection design analyzes for an intersection between Norra Promenaden and Packhusgatan in Norrkoping city, Sweden. Microsimulation models are constructed. After obtaining a calibrated and validated simulation model of the current traffic situation, a signalized intersection, a roundabout and a “bowtie” design is proposed and evaluated in terms of travel time passing the intersection. The area studied, an intersection between Norra Promenaden/Packhusgatan in Norrkoping, Sweden, is a 4-leg intersection with 2 in each arriving approach. See Figure 1 for an overview of the intersection. The current design is an unsignalized roundabout with give way signs at the approach of each leg. The intersection is congested, especially during the afternoon rush hour.

Figure 1 The Norra Promenaden/Packhusgatan intersection

In this project, the traffic flows in each direction approaching the intersection is measured and used as input to the simulation. In order to perform the calibration and validation with different types of data, travel time in legs and queue lengths are chosen to be the data for calibration and travel time passing roundabout is chosen to be the data for validation.

The structure of the paper is as follows. The data collection procedure is described in Section 2. In Section 3, the calibration of the model is discussed and the calibration results are shown. In Section 4 the results from a model validation is presented, and in Section 5 models for alternative designs are presented. The paper ends with the conclusions in Section 6.

2. DATA COLLECTION

To build the simulation model, data have to be collected. The quality of the data is important since it influences the quality of the Aimsun model as well as the calibration, validation and improvement in the later part. In this section, we describe how traffic flows, travel times, and queue lengths are measured.

2.1 Traffic Flow

During the collection of traffic flows, the flow was measured every 15 minutes in weekdays. After observing the traffic flow for two days, the data shows that from 15:45 to 17:00, the traffic flows for each leg in each 15 minutes were very high compared to other time period. Hence, the peak hours have been identified to be from 15:45 to 17:00 in weekdays. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

The method used to measure the traffic flow is the following. First of all, the type of vehicle was separated into four categories: car, bus, truck and lorry. Secondly, the turning direction of each leg was divided into three directions: go straight, turning right and turning left. The turning proportion of one leg was defined in the formula (1).

. / / = (1) .

To measure the traffic flow from one leg to the other three legs, one person stands at the corner of the roundabout and counts number of each type of vehicles passing through the yield line of the intersection from 15:45 to 17:00. Each leg was measured five times and the mean value of traffic flows were calculated for the OD matrix creation as the input data for the Aimsun model. The mean values of the traffic flows for the different vehicle types, during the one hour and fifteen minutes time period, are shown in Figure 2. The real values in the figure are the mean values of 5 times counted traffic flow from each leg of the intersection. For instance, the value 1147.2 in the low middle of the Figure 2 shows that there were, on average, 1147.2 cars traveling from South leg to the north leg of the intersection during the studied time period.

Figure 2 Mean traffic flow from 15:45 to 17:00 in number of vehicles. At each turn, the first row represents the number of cars, the second trucks, third buses, and the fourth row number of lorries.

2.2 Travel Time on Legs

The travel time in one leg is the time for a vehicle passing through this leg. One of the legs measured, the travel time from point A to point B on the west leg, is shown in Figure 3. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

End point

Start point B

A Figure 3 Measurement of traffic time in legs

The travel times have been measured over a stretch of 100 meters for each leg, starting from the give way sign, which was located at the border of the leg and the roundabout. The north leg is an exception with the length of only 96.3 meters since there is a flyover limits the length to 96.3 meters (see Figure 5). Every 1.5 minutes, a car was chosen randomly at the start point with its front end passing the line, and group member A raised his/her hand, group member B at the end point saw the sign and started the timepiece. When the front end of that car passing the end point measured by group member B, the timepiece was stopped and the time was measured. Each leg was measured 3 times and totally 6 days spent on measuring the travel time in legs. For each time, the time period was from 15:45 to 17:00, with 50 samples every collection day, there were 150 travel time samples from each leg. Since the travel time data collections were carried out at the same period as the traffic flow counts, the number of samples was independent with the traffic flow from each leg. The collected values from the 3 observation days have been averaged to get 50 samples for use in the calibration procedure. The mean value for each leg is shown in Table 1.

Table 1 Measurement result of travel time in legs (secs) South West East North 19.23 30.17 41.85 29.77

2.3 Queue Lengths

The queue length is the number of cars stopped before the yield line waiting to travel into the roundabout. The queue lengths was measured at the same time as the measuring of the travel times. The length is measured in the unit number of cars. The group member A (see Figure 3) at the start point of travel time measurement also measured the queue length. Every 1 minute, the number of cars in the queue for the two lanes was counted, with the start point to be the give way sign. Each leg was measured 3 times and totally 6 days spent on measuring the queue length. For each time, the time period was from 15:45 to 17:00, with 75 samples have been taken every collection day, there were 225 queue length samples from each leg. The mean values are shown in Table 2.

Table 2 Measurement results of queue length (cars) South West East North 3.84 2.17 4.1 6.47

2.4 Travel Time Passing Roundabout

The travel time for passing the roundabout is measured as the traveling time of a vehicle Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011 passing through a leg to the opposite leg. For the west, east and south legs, 100 meters from the give way sign were measured to be the start point A (see Figure 4). Only the travel time for cars going straight ahead was measured, the give way sign of the opposite part was set to be the end point B. Every 1.5 minutes, the travel time was measured once using the same method used for measuring travel time, the travel time measured is valid from the beginning of a section to the beginning of the opposite section. The start (A) and end (B) points for the west leg are shown in Figure 4.

End point

B Start point

A Figure 4 Measurement method of traffic time passing roundabout for west, east and south parts

For the north direction, 96 meters from the give way sign were also used as the start point, which aimed to match the model. However, the group member C at the end point cannot see the group member A at the start point. One more person was required to stand at the middle of the roundabout. When the group member A at the start point raised his/her hand, the group member B who was in the middle started the timepiece. When the car passing group member B, B raised his/her hand to show, which car were being measured to the group member C who was standing at the end point. When the car passed the end point, the group member C raised his/her hand. The middle one B stopped the timepiece and wrote down the time. The time interval was also 1.5 minutes. The measurement points used are shown in Figure 5.

Figure 5 Measurement method of traffic time passing roundabout for north parts

Each leg was measured two times and two days was spent on measuring the travel time passing the roundabout. For each time, the time period was from 15:45 to 17:00, with 50 samples have been taken every collection day, there were 100 travel time samples from each leg. The number of samples was independent with the traffic flow from each leg. The mean Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011 values are shown in Table 3.

Table 3 Measurement result of travel time in roundabout (secs) S-N W-E E-W N-S 28.44 44.33 56.02 45.35

3. CALIBRATION

The calibration process began with the input of some fixed parameters such as vehicle length, car width, etc. The more effective parameters then began to be adjusted to provide a relatively good fit between the observed and simulated data. At last, some parameters with smaller impact are calibrated for optimal modeling results. According to Burghout (2004), there is no exact number of replications that must be made. The most common number of replication used is between 5 to 10 replications. This study uses 10 replications in the experiments.

3.1 Calibration Procedure

The vehicle’s local vehicle parameters in Aimsun are set once and kept unchanged until the calibration procedure was finished. Those parameters are: 1. Vehicle length 2. Car width 3. Minimum distance vehicle 4. Sensitivity factor

The parameters in the Aimsun model determine the three basic vehicle behavior models (gap acceptance, car following and changing) (Aimsun 6 User’s Manual, 2009). In this model, the gap acceptance model is the most significantly influence to the vehicles behavior since every vehicles have to stop at the yield line (at the give way sign) waiting for a gap to travel into the roundabout. The following parameters were calibrated during the main calibration procedure: maximum desired speed, maximum acceleration, normal deceleration, speed acceptance and maximum give way time.

The experiments made in the calibration process were carried out with one change each run, starting from the default value in Aimsun. Because the default values was preset for highway simulations, the parameters typically need to be modified to make the vehicles move much slower to fit in the observed travel times and queue lengths. For most of the parameters, the exception being the give way time parameter, the calibration procedure will result in reduced values, which indicates a more patient travelling behavior than the default.

1. To change the maximum desired speed of each type of vehicle. 2. To change the maximum acceleration. At first the default value 3 was tested. The model gave very high mean speeds, therefore the maximum acceleration was reduced to get lower mean speeds. 3. To change normal and maximum deceleration. The default values of 4 and 6 were reduced. The cars’ parameters experienced a dramatically decrease, since the drivers in Sweden tend to start reducing their speed from further distance when they see a long queue length at the yield line. 4. To change speed acceptance. The default speed acceptance was 1.1. The experiments Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

made the values dropped to around 0.6 for all of the vehicles since most of the observed vehicles moved much slower speed than the speed limit due to the high volume traffic. 5. To change give way time. The roundabout has a give way sign in the each leg, so that it was imperative to consider the maximum give way time. The default value of 10 sec was increased to 19 sec for cars and 60 sec for other vehicles since the driver could wait for more time in case of a roundabout than in the normal highway case.

Aimsun has a parameter for each turning section called automatic speed. The automatic speed check box relates to the speed of turning, the uncheck box means that vehicles’ speed could be forced around a certain value; otherwise the Aimsun will automatically set the speed. Since the north leg has a relatively high slope, which dramatically reduced the vehicles’ velocity from this direction and increased the travel time (see Table 2 and Table 4), all the three turns from this leg were set to use the option automatic speed.

The lane changing model does not have large impact on the vehicle modeling in this case, since the length of each leg is just equal to or smaller than 100 meter, and every leg has exactly 2 lanes, without any turning lanes near the roundabout. Hence, most of the vehicles are already at their desired lane before entering the model. Observations of the traffic at the intersection showed that most of the vehicles changed to their desired lane before 100 meters to the roundabout. Some of the local parameters which have high influence on the lane- changing model such as distance zone 1, distance zone 2, time distance on-ramp therefore were set to the default Aimsun values. Other local parameters were also set to default, since they had insignificant impact to the modeling results: section speed limit, lane speed limit, turning speed, visibility distance at intersections, yellow box speed, maximum give way time variability, reaction time variation, jam density, and the reaction time factor. The final parameters of the car, bus, truck and lorry gained from calibration are listed in the figures 6, 7, 8, and 9.

Figure 6 Car attributes setting Figure 7 Bus attributes setting Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

Figure 8 Truck attributes setting Figure 9 Lorry attributes setting

The steps for the calibration process are explained below. Assume that x is the array of the observation data and y is the array for the simulation result. The mean value of the observation and simulation data are calculated as:

= ∑ and = ∑ (2)

The standard deviation of observed data, and the standard deviation of model output, is calculated as:

= ∑ ( − ) (3)

= ∑ ( − ) (4)

To decide whether the observed data was the same with the simulated data or not, we compare and . We construct the following hypothesis:

: = (5)

: ≠ (6)

∝ If | | < ( ) the hypothesis : = cannot be rejected, otherwise the calibration ∝ process will be continued by adjusting parameters. The value is the t-table value, for + − 2 degrees of freedom, where is the number of observed data points and is the number of simulation data points.

. = (7)

( ) ( ) = (8) Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

The calibration in this model uses the confidence level of 95% (α=0.05).

3.2 Simulation Results

The car travel times from the simulation is shown in Table 4 as the average value of ten replications, while the queue length is the average value observed in the animated simulation every 1 minute.

Table 4 Travel time in the section (secs) and queue length (cars) simulation results West North East South Travel time (secs) 30.17 29.77 41.85 19.23 Queue length (cars) 2.17 8.47 4.10 3.84

The car travel times for passing the roundabout, shown in Table 5, will be used as the validation data.

Table 5 Travel time inside roundabouts (secs) West to East North to south East to west South to north 44.33 45.34 6.14 28.44

The OD-flow in the output of Aimsun, shown in Table 6, is checked to ensure that the model replicates the input OD-flow.

Table 6 The Aimsun output OD-flow (cars) West North East South Model 430.75 1307.5 663.25 1521.75 Observed flow (mean value) 383.4 1249.8 605.4 1431 Differences (%) 10.99 4.41 8.72 5.96

3.3 Travel Time Calibration Results

In Table 7, the observed travel times and the outputs from the Aimsun simulation is shown. ∝ The ( ) value for travel time is 2.0018 with = 50 and = 10.

Table 7 Observed travel time, output from Aimsun model and the calculation results (secs) Observed travel time West North East South n 50 50 50 50 x 30.17 29.774 41.852 19.229 S 9.72 9.86 22.24 6.13

Aimsun model West North East South n 10 10 10 10 y 27.24 31.57 44.88 16.86 S 2.25 3.93 2.25 0.28

84.823 88.315 444.791 33.545 S 9.21 9.40 21.09 5.79 |T| 1.18 0.55 0.41 1.18 t-table 2.0018 2.0018 2.0018 2.0018 Valid yes yes yes yes Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

∝ From Table 7 we conclude that | | < ( ) holds, which implies that the hypothesis : = cannot be rejected.

3.4 Queue Length Calibration

The second type of data used in the calibration process is the queue length. The observed queue length was collected every 1 minute for the two lanes. A similar method was used to collect queue lengths from the simulation; the simulation run provides the number of car in queue every minute. A similar process as for the travel time calibration is applied. The ∝ hypothesis : = cannot be rejected if | | < ( ). The t-table value for =75 and = 10 is 1.982827. Table 8 shows the observed and simulated queue lengths.

Table 8 Observed queue length and output from Aimsun model (cars) Observed queue length West North East South 75 75 75 75 2.17 8.47 4.10 3.84 1.2 2.48 2.4 2.21 Aimsun model West North East South 10 10 10 10 1.68 5.36 4.5 3.026 1.7 5.36 4.55 3.26

2.47 17.43 13.25 7.77

1.57 4.18 3.64 2.79 | | 1.90 1.88 0.28 1.78 t-table 1.9828 1.9828 1.9828 1.9828 Valid yes yes yes yes

The | | values for the four legs is smaller than threshold t-table value, that is, | |<1.9828, which implies that the hypothesis : = cannot be rejected.

4. VALIDATION

The validation process aims to determine whether the constructed model is acceptable or not. An acceptable model means that the model can represent the reality accurately enough. The observed data not used in the calibration process is used for the validation, that is, the travel time for passing the roundabout is used for the validation. The tests used in the calibration process can be applied also for the validation process. The x represents the observed data and ∝ y represents the simulation result. The ( )value for west leg is 2.0357 and for the others 2.0018. The observed travel times for the four legs and the corresponding simulation outputs are shown in Table 9.

Table 9 Observed car travel time passing roundabout and output from Aimsun model (secs) West to east North to south East to west South to north 25 50 50 50 44.33 45.35 61.14 28.45 12.53 20.26 29.29 8.82 ±5.17 ±5.76 ±8.40 ±2.51

Aimsun model Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

West North East South 10 10 10 10 43.11 43.02 68.95 25.91 4.45 4.09 6.13 0.29 ±3.19 ±2.93 ±4.38 ±0.21

119.51 349.35 745.60 65.74 10.93 18.69 27.31 8.11 | | 0.30 0.36 0.83 0.90 t-table 2.0357 2.0018 2.0018 2.0018 Valid Yes yes yes yes

From Table 9, we observe the at all | | values are smaller than 2.0357 and 2.0018, indicating that, with the 95% level of confidence, the hypothesis of that the average observed data is equal to the average simulated data output cannot be rejected.

5. EVALUATION OF ROUNDABOUT AND SIGNALIZED INTERSECTION

Currently, the intersection design is a roundabout. The traffic simulation model reflects the real traffic sufficiently well, as stated in the previous sections of this paper. Hence, by simulating alternative designs it is possible to compare, for example, two designs: roundabout and signalized intersection. To test out alternative designs for the Norra Promenaden/ Packhusgatan intersection, alternative models have been constructed. The travel time passing the intersection has been chosen as the evaluation measure for the comparison between each alternative.

Firstly, a normal signalized intersection design model has been built. According to the Highway Capacity Manual (2000), the principle of finding the optimal green time for each phase is made by calibration. After choosing initial green durations for each phase, different green time combinations have been tested in the Aimsun model to find the best possible combination. Because the turning left proportion in every direction is high, according to the Highway Capacity Manual (2000), this intersection may need 4 phases, with 2 phases for turning left traffic. The four phase design is shown in Figure 10.

Figure 10 Four phase of the signalized design (FHWA, 2009)

The duration of green time for each phase is calculated based on the flow ratio. The yellow time has been set to tZ = 3s, which is the Aimsun default value. The total cycle time is limited to 130 seconds. The initial green time durations for the four phases is shown in Table 10. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

Table 10 Initial duration of green time for each phase Phase 1 Phase 2 Phase 3 Phase 4 Maximum flow 1016 351 129 133 Duration (s) 75 26 10 10

Around 20 combinations of green time have been tested within the model in order to reduce the travel time passing the roundabout. The combinations has been tested one by one in a way similar to a calibration process, with the aim to find a combination which gave the smallest travel time for passing the intersection. The final green time durations for the four phases is shown in Table 11.

Table 11 Duration of green time for each phase Phase 1 Phase 2 Phase 3 Phase 4 Duration (s) 52 35 20 20

No pedestrian movements are taken into account in this calculation. The results for running the simulation model with 4 phases signal and the same parameters as the original design is shown in Table 12.

Table 12 Comparison of travel time passing the intersection between signalized intersection alternative and original design West to South to North to East to east north south west Original 43.11 25.91 43.02 68.95 Signalized 72.51 160.72 137.43 289.77 Reduction -68% -520% -219% -320%

The travel time passing through the intersection has dramatically increased compared to the original design, especially in the Eastbound. Even without any hypothesis test, it can be concluded that this signalized design is not better than the original roundabout design. Therefore, the NCHRP Report 572 (2007) and the FHWA studies (FHWA-RD-00-067, 2000) are reasonable in this particular situation. In conclusion, the normal signalized intersection is not suitable for this particular intersection.

A mixed solution between the roundabout and signalized intersection is the “bowtie” intersection. An overview of this intersection type is shown in Figure 11. The idea is to use two roundabouts placed at east and west, away from the main intersection, to accommodate left turns and allow more traffic to flow comfortably through the main intersection (Highway Capacity Manual, 2000). Hence, inside the main intersection, left turning is prohibited.

Figure 11 The bowtie intersection (Wilbur Smith Associates, 2008)

Bowtie intersections are believed to accommodate more vehicles per hour than a normal signalized intersection or roundabout (Wilbur Smith Associates, 2008). However, for this particular intersection, we will use simulation to evaluate its efficiency. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

To analyze the impact of applying a bowtie intersection to the studied roundabout, the same network layout and parameters as for the signalized intersection are used. The two supporting roundabout at the East and West of the intersection are not included in the model, but will have an impact on the traffic situation. Turning left inside the intersection will be prohibited, all the left turning flows will be added to the right turning flows and straight head for adequate directions. Inside the roundabout, signalized control is used to control the traffic.

Figure 12 The bowtie intersection simulation model

A new OD-matrix is created as input to the Aimsun model. Because vehicles which want to turn left must turn right and make a U-turn at one of the supporting roundabouts at west or east direction. This alternative needs only two phases as control phases, they are chosen as shown in Figure 13. The green time for each phase has been calculated and tested with same the method used in normal signalized intersection.

Figure 13 The bowtie intersection signal phases (FHWA, 2009)

Table 13 Duration of green time for each phase Phase 1 Phase 2 Duration (s) 52 35

Finally, Aimsun simulation provides the travel time for passing through the intersection Norra Promenaden/ Packhusgatan in the bowtie design. The travel times are given in Table 14.

Table 14 Comparison of travel time passing the intersection between alternative 3 and original design West to South to North to East to east north south west Original 43.11 25.91 43.02 68.95 Bowtie 44.34 36.01 36.87 56.22 Reduction -1% -39% 30% 27% Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011

The Aimsun results show a more equally distributed travel time for each direction of the intersection, compare to the original design. Vehicles can travel faster from the west, north and east of the intersection, while there is an increase in travel time for vehicles in the southbound direction. The hypothesis test has been used to find the impact of these changes.

Table 15 Hypothesis test of the travel time passing the intersection, alternative 3 compared to original design West to east South to North to East to north south west 4.45 5.37 4.09 6.13 1.80 10.69 6.52 14.31 11.53 71.51 29.60 121.15 3.40 8.46 5.44 11.01 T -0.81 -2.67 2.53 2.59 T 2.101 2.101 2.101 2.101 Impact Insignificant Significant Significant Significant

The hypothesis test results indicate that while the improvements of travel time in the northbound and eastbound were significant; the increase of travel time in the southbound is significant. The change of travel time in westbound direction is considered to be nondramatic. Because it is difficult to find the impact of these changes on the overall intersection, a hypothesis test will be conducted with the total travel time of all the 4 legs. The data for this test is given in Table 16. Table 16 Hypothesis test of total travel time passing the intersection in 4 legs, alternative 3 compared to original design Total travel time 9.31

5.68

59.44

7.71

T 2.19

T 2.101 Impact Significant

The total travel time of the 4 legs in this alternative equal to 169.0 seconds, compared to 181.0 seconds of the original design. This decrease in total travel time passing through the intersection was also proven to be significant in the hypothesis test. Therefore, this alternative is a significant improvement to the intersection, in terms of measured the travel time.

Travel time for passing the intersection has been chosen as the measure for comparison the bowtie design with the original one. However, the travel time for evaluation is only measured for the traffic going straight head. Hence, it is only possible to conclude that the bowtie design has a better performance in terms of travel time passing the roundabout compared to the original design for the going straight traffic. This is the majority of the traffic flow in this case.

6. CONCLUSION

In this report results from a microsimulation of the real traffic situation during weekdays Proceedings of the Eastern Asia Society for Transportation Studies, Vol.8, 2011 afternoon rush hours is presented. The simulation model has been calibrated using travel time and queue lengths in each leg, and validated using travel time passing the intersection. The real traffic data in form of vehicles flows; travel time and queue lengths were observed during afternoon rush hours (15:45 to 17:00) and used as input to the model.

The results from the simulation study indicate that the roundabout design provides a better performance in terms of travel time compared to a traditional signalized intersection design. A hybrid version of a roundabout and a signalized intersection – a bowtie intersection shows a significantly better performance than both roundabout and signalized design (for the going straight vehicles). This project has been carried out with a “standard” type of intersection with large clearance, where passenger cars are the dominating vehicle type (over 90%).

The analyzing results based on a standard type of intersection as the roundabout in Norrkoping city, Sweden gave initial suggestions for a reasonable traffic control at congested intersections in big cities. The results of microsimulation and the performance of different design alternatives could be a good reference for future researches on similar intersections with high traffic volume and car dominated traffic.

ACKNOWLEDGMENT

We would like to acknowledge MSc. Sara Respati and Ms. Yi Qian, for their work with the data collection and the writing of an earlier version of this report.

REFERENCES

Aimsun 6 User’s Manual (2009) TSS-Transport Simulation Systems. Alisoglu, S. (2010) Roundabouts v. signalized intersections: A comprehensive analysis, Kansas Government Journal, July 2010. Burghout, W (2004) A note on the number of replication runs in stochastic traffic simulation models. http://www.ctr.kth.se/publications/ctr2004_01.pdf. Retrieved March 19, 2011. Highway Capacity Manual (2000) Transportation research board. NCHRP Report 572: Roundabouts in the United States (2007) National Cooperative Highway Research Program, TRB, NAS, Washington DC. FHWA (2000) Roundabouts: An Informational Guide, FHWA-RD-00-067, U.S Federal Highway Administration. Shokri, F. Mokhatarian, H., Wasmail, A. and R. Rahmat R. (2010) Comparing the Design of Roundabout and Intersection with aaSIDRA Software, European Journal of Scientific Research, Vol. 40, Number 2, 239-246. Traffic Signal Timing Manual, Chapter 4, Traffic Signal Design (2009) FHWA http://ops.fhwa.dot.gov/publications/fhwahop08024/chapter4.htm. Retrieved March 19, 2011. Younes, B. (2000) Roundabouts vs. Intersections: The Tale of Three UAE Cities, A review of historic developments and current practice with an aerial photographic documentation. ITE Annual Meeting Compendium, Washington (DC): Institute of Transportation Engineers http://www.ite.org/traffic/documents/AB00H5001.pdf. Retrieved March 19, 2011. Wilbur Smith Associates (2008) Innovative Intersections: Overview and Implementation Guidelines, Compass: Community Planning Association of Southwest Idaho.