A simulation model for the mixed traffic system in Vietnam

Int. J. Simulation and Process Modelling, Vol. 5, No. 3, 2009 233 A simulation model for the mixed traffic system in Vietnam Quynh-Lam Le Ngoc*, Ngoc-Hien Do and Ki-Chan Nam Department of Logistics Engineering, Korea Maritime University, #1 Dongsam-dong, Yeongdo-gu, Busan, 606-791, Republic of Korea E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] *Corresponding author Abstract: Simulation models have been used successfully to model traffic systems, test various traffic control algorithms and solve traffic problems in many developed countries. However, they are unlikely to produce reliable results if applied to Vietnam’s mixed traffic conditions, where motorbikes are the principle means of transportation. This paper models and simulates the mixed traffic system. Accordingly, the traffic system is characterised and the logic of the simulation model is introduced briefly. In order to validate and demonstrate the usefulness of the simulation model, a case study is presented. Finally, some conclusions and suggestions are proposed. Keywords: simulation model; mixed traffic system; traffic problem. Reference to this paper should be made as follows: Le Ngoc, Q-L., Do, N-H. and Nam, K-C. (2009) ‘A simulation model for the mixed traffic system in Vietnam’, Int. J. Simulation and Process Modelling, Vol. 5, No. 3, pp.233–240. Biographical notes: Quynh-Lam Le Ngoc is a Doctoral student at the Department of Logistics Engineering, Korea Maritime University, Republic of Korea. She received her BEng in Electronics Engineering from the Ho Chi Minh City University of Technology (HCMUT), Vietnam, and her MEng in Industrial Systems Engineering from the Asian Institute of Technology (AIT), Thailand. She is a Lecturer at the Industrial Systems Engineering Department, HCMUT. Ngoc-Hien Do is a Doctoral student at the Department of Logistics Engineering, Korea Maritime University, Republic of Korea. He received his BEng in Industrial Systems Engineering from the Ho Chi Minh City University of Technology (HCMUT), Vietnam, and his MSc in Logistics Engineering from the Korea Maritime University, Korea. He works as a Lecturer and Researcher at the Department of Industrial Systems Engineering Department, HCMUT. His current research interests are simulation, scheduling, planning and multimodal transportation. Ki-Chan Nam is a Professor at the Department of Logistics Engineering, Korea Maritime University, Republic of Korea, and the Director of the New University for Region Innovation (NURI) project. He is also a member of a committee in the Busan Port Authority (BPA). He received his BA in Navigation from the Korea Maritime University, and his MSc and PhD in Transportation Planning from the University of Wales, UK. 1 Introduction Traffic congestion is a difficult problem that has attracted a great deal of attention in the Socialist Republic of Vietnam (hereafter, Vietnam). According to the Vietnamese Traffic and Public Works Service (Phan and Kien, 2007), approximately USD 875 million was lost in that year because of pollution, stress, productivity loss, time loss and delivery delays resulting from traffic congestion in HoChiMinh city, the largest and most dynamic city in the entire country. Efforts to solve this problem have included Copyright © 2009 Inderscience Enterprises Ltd. assigning different starting times for companies, universities and schools, reorganising traffic flows and developing infrastructure. Despite all these efforts, the situation has not been improved, principally because other alternatives were not evaluated before the efforts were applied. Suggested alternatives should focus not only on quantity, but also on quality. Therefore, any potential solution must be validated before being applied to a practical system to save time, money and effort. Much research concerning traffic and transportation problems has been conducted. Wen (2008) recommended 234 Q-L. Le Ngoc et al. a framework for a dynamic and automatic traffic light control expert system combined with a simulation model to help analyse the traffic problem. The model adopts inter-arrival and inter-departure times to simulate the number of cars entering and exiting the roads. The development, structure and evaluation of a road traffic control system simulator by Parallel Inference Machine (PIM) were proposed by Takahashi et al. (2002). Their research provided an effective description of a microscopic traffic model of urban districts and the analysis and problem solving of traffic congestion based on actual data. Herty et al. (2007) suggested models for vehicular traffic flow based on partial differential equations and their extensions to road networks. A fluid dynamic traffic model was simplified and a new traffic flow model was derived based on Ordinary Differential Equations (ODEs). Optimal control problems controlled by the ODE model were considered and the optimal system was derived. Many well-known simulation models are available for commercial usage such as QUADSTONE PARAMICS, VISSIM and TSIS-CORSIM (Alexiadis et al., 2007). QUADSTONE PARAMICS is a modular suite of a microscopic simulation tool that is designed to handle a wide range of scenarios from a single intersection through a congested freeway to an entire city’s traffic system. VISSIM is a tool to simulate different traffic scenarios before starting implementation and it can simulate urban and highway traffic, including pedestrians, cyclists and motorised vehicles. TSIS-CORSIM is another microscopic traffic simulation software package for signal systems, freeway systems, or combined signal and freeway systems. Vehicular movements in these models are achieved through car-following and lane-change models. Although the aforementioned models have been applied successfully in traffic systems in many developed countries such as the USA, Canada and Australia, they are unlikely to give reliable results if applied to the traffic system in Vietnam owing to the presence of distinct characteristics that make it different from the more homogeneous traffic systems of other countries. These characteristics will be described clearly in the next section. Therefore, it is essential to develop a simulation model appropriate for such a traffic system. Accordingly, the traffic problem in Vietnam is outlined in Section 2, and the logic of the simulation model is described in Section 3. A case study is proposed in Section 4 for validation and illustration. Conclusions are presented in Section 5. 2 In terms of the transportation infrastructure and superstructure, the urban road systems comprise long, narrow and interlacing roads of poor quality. Generally, there are no dedicated paths for pedestrians and cyclists. Therefore, all road participants have to negotiate the traffic. Traffic control systems are operated by traffic regulations, traffic light systems and policemen. There are no 24-hour automatic detective devices available for traffic supervision, which hinders a strictly regulation-observed traffic system. In addition, the traffic system in Vietnam is a mixed one with the traffic flow comprising both motorised, predominantly motorbikes, and non-motorised vehicles. According to the General Statistics Office of Vietnam (2008), about 21.72 million motorbikes were registered in 2007, in a population of 81 million, compared with only 1.11 million cars. The widespread use of the motorbike has arisen due to its flexibility and economy, the citizens’ income and road conditions. Despite the use of public transportation such as buses and taxis, they cannot compete with motorbikes in terms of flexibility, convenience in the narrow urban roads, time savings and low cost. The number of privately owned cars has increased continuously in recent years but only within the limitations imposed by high price, taxes and traffic congestion. Trucks and heavy trucks are used for freight transport because cities are normally economic centres where industries parks and ports are also located. In addition, the traffic system also comprises other transport modes such as bicycles and tricycles. The DVU’s behaviours are determined by the mixed traffic and existing infra- and superstructures conditions. Motorbikes may occupy any lateral position across the carriageways instead of travelling within a particular lane. To overtake each other, with traffic flowing on the right-hand side, cars have to change to the next lane on the left-hand side, whereas motorbikes have many choices, as illustrated in Figure 1. They may move one or two lanes to the left or right. Although moving to the right side or changing more than one lane to overtake another is illegal, it is ubiquitous. Priorities of lane usage are denoted from one to four, in decreasing order of priority. When DVUs want to leave the system or take the first exit at an intersection, they move towards the right. If they want to travel at high speed or make a U-turn, they tend to use the farthest lane on the left, which is usually dedicated for cars. Figure 1 Flex-passing rules for motorbikes (see online version for colours) Outlines of the practical traffic system Many unique characteristics support the differentiation of the traffic system in Vietnam from that of other countries. The most distinctive characteristics are the transportation infrastructure and superstructure, Driver-Vehicle-Units (DVUs) and their behaviours and traffic flow. The DVU terminology is used to reflect the fact that movements are affected by both vehicular and driver characteristics (Quadstone, 2003a, 2003b). The traffic flow in Vietnam resembles an ant track. DVUs in the system exhibit flocking behaviours (Reynolds, 1987) in that individuals avoid collisions by driving at about the same speed as their neighbours. They tend to follow a A simulation model for the mixed traffic system in Vietnam leading DVU, thereby forming travelling groups on the road. The leader usually drives according to the free-flow, acceleration/deceleration manner. Furthermore, vehicles move at different velocities and accelerations/decelerations depending on the drivers, vehicles’ physical performances and traffic contexts. The system’s complexity is raised at intersections where traffic flows intersect with each other. All types of vehicles mix while following chaos-rules, i.e., they are guided by traffic rules but travel in chaotic ways. As no vehicle has priority, they drive as they can. Some examples are illustrated in Figure 2. Figure 2 (a) Congestion on a road and (b) at an intersection (see online version for colours) (a) (b) These aforementioned characteristics create the distinction between the mixed traffic system and the lane-based, motorised traffic systems of developed countries. Because of these differences, research results based on the latter situation are unlikely to give reliable results if applied in Vietnam. Therefore, it is essential to develop a simulation model appropriate for the traffic system in Vietnam, in which all distinct characteristics are well imitated. 3 235 Logic of the simulation model A traffic system consists of a road network, DVUs and traffic control system. For the road network, which includes nodes (intersections) and arcs (roads) connecting nodes, a sub-lane notion is used for two-wheeled vehicles. The road network’s attributes such as the numbers of nodes and arcs, their sizes and the numbers of lanes and sub-lanes in each road are considered in the simulation model. To reflect the DVU’s behaviours, the physical, dynamic and other relevant characteristics are also considered. The physical characteristics comprise the types and sizes of vehicles, whereas the dynamic ones consist of maximum and minimum velocities, and minimum, mean, and maximum accelerations/decelerations of vehicles. The traffic control system at intersections includes a traffic light system and traffic circle. To imitate the specific characteristics of the traffic system in Vietnam, many techniques and logic models have been applied. A vehicle referring technique based on a three-dimensional coordinate system has been used to model the vehicles’ locations on the mixed traffic road. Since vehicles may occupy any lateral position across the road’s width, this technique can accurately represent the vehicle’s movements. Vehicles are symbolised by coloured rectangles of size corresponding to their real sizes. If two or more types of vehicles have the same size but different dynamic attributes, they are shown in different colours. Their locations are determined by (x, y, z) coordinates, in which the (x, y) coordinates describe the locations of vehicles on the same road, whereas the z-coordinate is used when an overpass system exists. Many logic models have been applied to simulate DVU behaviours, such as car-following, free acceleration, lanechanging, direction-changing, stop-run and intersectionconflict models. The simulation model assumes that the acceleration/deceleration of the following DVU depends on its current speed, control speed and the perception-reaction time of driver. The control speed is defined as the minimum value of vehicle’s maximum speed, lane/sub-lane maximum speed and driver’s desired speed. The perception-reaction time of the driver depends on the current spacing, desired spacing between it and lead vehicle immediately in front (local leader) and the status of its local leader(s). When moving on the road, if the distance from a DVU to a local leader is less than a predefined critical value (d0), it will follow its local leader with an acceleration/deceleration that is computed by equation (1), according to the car-following model. If the distance is more than the critical value, its movement is guided by the free acceleration/deceleration model as in equation (2). an (t ) = α vn (t − Tn ) β ∆vn (t − Tn ) ∆xn (t − Tn )γ ∆vn = vn -1 (t ) − vn (t )  ∆xn = xn -1 (t ) − xn (t ) (1) 236 Q-L. Le Ngoc et al. in which an(t) is the acceleration/deceleration of vehicle n at time t; α, β and γ are factors that are determined as shown in Table 1. Because the value of acceleration and that of deceleration are different, each coefficient has two values. Tn is the reacting time; vn(t) is the velocity of vehicle n at time t; xn(t) is the position of vehicle n at time t; ∆vn and ∆xn are the differences in velocity and distance, respectively, between the two vehicles. Table 1 Coefficients Coefficient α β γ Acceleration 9.21 –1.67 –0.88 +  amax, vn < vtarget , n n :  an =  0 : vn = vtarget, n  −  an , n : vn > vtarget , n Decelerate 15.24 1.09 1.66 (2) + is the maximum acceleration of vehicle n, in which amax,n vtarget,n is the control speed of vehicle n and an−, n is the normal acceleration of vehicle n. Lane/sub-lane-changing behaviours can occur when a DVU wants to move to another lane/sub-lane that allows its desired speed, or dodge obstacles, or change to a suitable lane/sub-lane before changing its direction. The lane-changing model can imitate these behaviours via a modification that enables it to simulate the different lane-changing behaviours of different types of vehicles as mentioned in the previous section. Before changing lane/sub-lane, DVUs have to look for suitable gaps. Having found such a gap, a DVU will accelerate without exceeding the maximum safe speed concerning new lead and follow vehicles. Before entering an intersection, a DVU has to check the traffic lights and change to a suitable lane/sub-lane to change its direction. The consecutive direction of a DVU such as turn-left, turn-right, or straight ahead follows a direction-changing distribution that is observed and collected from the field. Changing-lane/sub-lane activity is implemented when a DVU remains at a stipulated distance from the intersection. In the case of a green traffic light, the DVU will enter the intersection; otherwise, it has to decelerate to stop at the stop-line or join the queue. A stop-run logic model is applied to simulate these activities. The intersection-entering behaviours of DVUs differ markedly depending on the presence or absence of a traffic circle. In its absence, suitable intersection-conflict models are used, in which different priorities are assigned for DVUs concerning the directions that DVUs are entering and exiting. In its presence, if the DVU takes the first exit from the traffic circle, it usually moves towards the right, otherwise, it approaches the traffic circle. However, these aforementioned techniques and logic models are only representative ones; other techniques and logic models were also applied to simulate the characteristics of the mixed traffic system. A simulation programme was developed using Visual Basic 6.0 as a programming language. The programme comprises three sub-programmes, RoadStruct, Sim and Ani as shown in Figure 3. The main function of RoadStruct-programme is to convert data collected from the field into simulation data used in Sim-programme. Sim-programme is the core of the simulation programme, as it simulates the behaviours of the traffic system. The obtained results are recorded on two different formatted files. One is used to make reports, which are the user’s expected factors, whereas the other is an animation file used as input data for the next programme, Ani-Programme. This programme supports the analysts in observing or analysing suggested alternatives or scenarios through the animation displays. Figure 3 Logic of the simulation programme Some factors or indexes such as Volume IN, Volume OUT, Average Speed and Density Index are used to validate the simulation model and evaluate the simulation results. These factors are recorded directly or indirectly through parameters obtained from counter machines set-up at all road inputs and outputs. Among these factors, Volume IN (VIij – vehicles/minute) is the average number of vehicles of type i entering the system on road j, Volume OUT (VOs– vehicles/minute) is the average number of vehicles of type i exiting the system on road j and the Average Speed is determined through the vehicles’ travelled distance and time. The density of traffic is determined by using a formula provided by Gazis (2002). 4 Validation and application The applications for this programme are diverse. It can be used to determine a system’s capacity or ameliorate traffic congestion problems at a traffic node or a network. Proposed alternatives include adding or removing traffic circles or a road section. The assigning of traffic flows is also simulated before being applied. It can also be used to determine the optimal traffic light system. This section presents a validation example in which a traffic problem is depicted, followed by simulation results and discussion. 4.1 Problem description A traffic node, comprising the intersection of seven roads, as shown in Figure 4, is considered. A traffic circle A simulation model for the mixed traffic system in Vietnam regulates the traffic flows. A traffic light system was installed to control the traffic flows. Since this node is formed by major routes and surrounded by many schools, hospitals and office buildings, the rush-hour traffic is horrendous. The simulation programme was used as an efficient study tool to improve that situation. Figure 4 237 Figure 5 Simulation procedure Table 2 Mean of number of vehicles/minute entering the system Traffic intersection (see online version for colours) 4.2 Data collection and simulation scenarios To achieve the aforementioned objectives, a simulation procedure is illustrated as shown in Figure 5. A simulation model was constructed based on the objectives such as reorganising traffic flows and considering effects of the traffic circle or traffic light system, observations such as DVUs’ behaviours, traffic flows and structures of intersections, traffic theories mentioned in Section 3, and necessary data such as vehicles’ information, a map of road system, and time and organisation of the traffic light system. Suitable factors and techniques were used to validate the model, which was compared with the existing system. Evaluation indexes and alternatives were suggested by experts and simulated. Data analysis techniques were then used to evaluate the simulation results of the considered alternatives. Expected factors were determined to support the decision-making. The vehicle volumes entering and exiting the system were observed and collected at the beginning and ending points of each road, at different rush hour periods per day, over 30 consecutive days. Their means are shown in Tables 2 and 3. Six types of vehicles were considered on the seven roads of the system: bicycle, motorbike, car, small truck, truck and bus, numbered from 1 to 6, respectively. To determine the turn-ratios at the intersection and the average vehicle speeds in the system, the vehicles were traced through video tapes. The turn-ratios were computed as an average turn percentage by observing 50 vehicles per type each time for a total of six times, whereas the average speeds were obtained through the travelling distance and time as shown in Table 4. Vehicle type Road No. 1 2 3 4 5 6 2.2 74.72 4.13 3.35 0 0.42 II 2.36 70.3 6.05 5.66 2.5 0.38 III 0 0 0 0 0 I 0 IV 0.92 20.37 0.38 0.52 0 0.01 V 2.27 78.82 3.6 3.19 0 0.39 VI 2.88 82.03 5.45 4.8 0 0.4 VII 0.15 13.39 0.56 0.15 0 0 Table 3 Mean of number of vehicles/minute exiting the system Vehicle type Road No. 1 2 3 4 5 6 I 2.69 79.32 5.32 4.44 0 0.48 II 0 0 0 0 0 III 2.18 70.32 3.93 3.21 0 0.25 IV 0.63 23.44 0.62 1.6 0 0.02 V 2.08 78.05 4.1 3.43 0 0.34 VI 2.8 83.71 5.7 4.69 2.5 0.51 VII 0 0 0 0 0 Table 4 Average vehicle speeds in the system (km/h) 0 0 Vehicle type 1 2 3 4 5 6 6.66 29.05 24.75 20.94 4.04 9.94 238 Q-L. Le Ngoc et al. Among the 10 alternatives that we considered, the first is the do-nothing alternative, A-pre, as shown in Tables 5 and 6. The other nine consist of varying traffic light times, enlarging roadways and reversing the direction of a one-way street. Some are combinations of other alternatives. An example of a traffic lights system with two phases is illustrated in Figure 6. Figure 6 Table 5 Two-phase traffic light system (see online version for colours) Description of alternatives Alternative Description Table 7 A_pre Do–nothing alternative. Light times of two phases are as in Table 6, in which phase 1 concerns roads I, IV, V, and VII, and the remainder belong to phase 2 A_1 Similar to A-pre, except traffic light times varied as presented in Table 6 A_2 A_3 A_4 A_5 A_6 A_7 A_8 A_9 Table 6 including Volume IN, Volume OUT and Average Speed, were validated by testing the resemblance between the simulation results and actual observations. The simulation results of the do-nothing alternative, A-pre, were recorded in 30 min of simulation for a run. One-way Analysis of Variance (ANOVA) tests were applied to detect differences among means between the observed and simulated results of each validation factor. Although the ANOVA tests were conducted for each validation factor at each input or output road, only a few representative results were provided in this section, in which the Volume IN was tested for road II, the Volume OUT was tested for road VI and the Average Speed was tested for the whole system. The p-values of these tests are shown in Table 7. Changing traffic light phases and varying traffic light times, in which phase 1 includes road Nos. I, II, IV, and VII; the light times as in Table 6 Varying traffic lights times as in Table 6, and reversing direction of the one-way road (VII); the traffic light phases remain the same as previous alternatives Similar to A_7 and enlarging roadway at a corner as noted in Figure 4 Light times (s) of considered alternatives Phase 1 Phase 2 Alternative Green Yellow Red Green Yellow Red Clear A_pre 25 5 30 20 5 35 5 A_1 30 5 35 25 5 40 5 A_2 21 5 34 24 5 31 5 A_3 25 5 30 20 5 35 5 A_4 30 5 35 25 5 40 5 A_5 21 5 34 24 5 31 5 A_6 25 5 30 20 5 35 5 A_7 30 5 35 25 5 40 5 A_8 21 5 34 24 5 31 5 A_9 30 5 35 25 5 40 5 4.3 Model validation The simulation model was validated through many processes. The entities’ behaviours were observed via animation displays, whereas the system’s parameters, Level of significance (p-values) of ANOVA tests Vehicle type 1 2 3 4 5 6 ~1 Volume IN 0.861 0.518 0.883 0.888 0.932 Volume OUT 0.908 0.975 0.798 0.761 0.983 0.937 Average speed 0.691 0.759 0.115 0.681 0.200 0.172 In the test results, none of the validation variables exhibit any significant difference between the observed data and simulation results. This good fit supports the validity of the simulation model in simulating the traffic system under mixed traffic conditions. 4.4 Simulation results Although any decision to select a solution among the alternatives is beyond the scope of this study, the simulation model is a useful support tool for decision-makers to make judicious decisions, as the simulation results are a good point of reference. The proposed alternatives are simulated, and the simulation results are shown in Table 8, which presents four evaluation factors according to six types of vehicle for each alternative. For a simple comparison, the general value of each factor is calculated and summarised as shown in Table 9. The factor VI, the number of vehicles served by the system, is the sum of VIi, which shows the system’s capacity. The factor VO, the sum of VOi, presents the system’s productivity. The factor AS, computed by equation (3), indicates the system’s service speed. ∑ AS = 6 i =1 ASi ×VOi VO . (3) Finally, the factor D, the sum of Di, illustrates the system’s busy level. A simulation model for the mixed traffic system in Vietnam Table 8 Simulation results Table 9 Types of vehicle Alt. A_pre A_1 A_2 A_3 A_4 A_5 A_6 A_7 A_8 A_9 Factor 1 VIi 10.78 339.63 20.17 VOi 10.38 334.84 19.67 2 239 3 Alternative comparison Alt. 4 VI VO AS D A_pre 392.35 386.36 27.62 944.90 5 6 17.67 2.50 1.60 A_1 388.36 385.46 22.56 1125.05 17.37 2.50 1.60 A_2 385.66 384.46 22.59 1115.64 ASi 6.66 29.05 24.75 20.94 4.04 9.94 A_3 393.75 391.15 22.59 1117.86 Di 97.12 701.47 48.90 50.63 37.13 9.66 A_4 383.07 379.37 27.57 931.56 VIi 10.78 333.64 20.17 16.97 2.30 1.60 A_5 377.78 364.50 27.33 916.48 VOi 10.78 333.05 19.87 16.87 2.30 1.60 A_6 377.48 375.88 27.31 902.44 ASi 6.72 29.23 24.43 18.44 4.21 7.00 A_7 385.46 384.47 27.64 932.36 Di 96.25 684.86 49.54 55.22 32.78 13.71 A_8 382.57 381.17 22.37 1103.36 VIi 10.78 330.05 20.27 17.97 2.50 1.50 A_9 393.95 376.38 22.44 1141.65 VOi 10.38 327.65 19.97 17.57 2.30 1.50 3.99 8.42 ASi 6.73 29.23 22.66 19.25 Di 96.11 677.49 53.67 56.01 VIi 10.38 325.86 20.47 17.07 2.40 1.60 VOi 10.38 313.88 19.87 16.57 2.20 1.60 3.15 8.06 37.59 10.69 ASi 8.43 28.96 23.85 17.55 Di 73.88 675.12 51.50 58.36 VIi 10.78 325.26 19.87 17.47 2.60 1.50 VOi 10.78 324.06 19.77 17.27 2.50 1.50 4.66 8.01 45.71 11.91 ASi 8.41 28.87 22.90 19.86 Di 76.91 675.98 52.06 52.78 VIi 10.38 336.64 20.07 17.27 2.40 1.60 VOi 10.38 334.24 19.77 17.07 2.40 1.60 4.18 7.78 33.48 11.24 ASi 5.85 23.67 20.22 17.59 Di 106.46 853.33 59.55 58.91 VIi 10.78 334.24 19.37 17.37 VOi 10.78 333.04 19.37 17.37 2.30 1.60 ASi 6.73 23.62 20.06 19.47 3.02 7.20 34.45 12.34 2.30 1.60 Di 96.11 849.04 57.94 53.53 VIi 11.18 340.23 20.47 17.77 45.70 13.33 VOi 10.78 339.03 20.17 ASi 7.58 23.60 20.53 Di 88.50 864.99 59.82 58.71 37.13 8.71 VIi 9.98 342.63 19.67 17.67 2.40 1.60 VOi 9.58 326.46 19.27 17.27 2.20 1.60 ASi 6.73 23.43 20.41 18.59 3.03 7.45 2.50 1.60 17.27 2.30 1.60 18.16 4.04 11.02 Di 88.97 877.41 57.82 57.03 VIi 10.38 331.25 19.37 17.27 47.52 12.89 VOi 10.38 330.05 19.27 17.27 2.60 1.60 ASi 6.76 23.45 19.98 17.95 4.72 7.13 Di 92.13 847.55 58.17 57.73 2.70 1.60 34.32 13.46 VIi: Volume IN of vehicle type i (vehicles/minute). VOi: Volume OUT of vehicle type i (vehicles/minute). ASi: Average Speed of vehicle type i (km/h). Di: Density index of vehicle type i (vehicles/km). By considering the alternatives according to the individual factors, we found that A_9, A_3, and A_7 maximise the system’s capacity in the busiest situation, productivity and service speed, respectively. To decide which alternative should be used, decisionmakers need to consider the relative weights of each factor based on the specific circumstances. In such a situation, certain multi-criteria decision-making techniques are useful. 5 Conclusions This paper presents the distinctive characteristics of a mixed traffic system and constructs an appropriate simulation model to simulate such a traffic system. Following validation through an application example, the model can be considered as a useful tool to make judicious decisions concerning traffic planning, traffic congestion solving and other traffic problems requiring solution in developing countries. References Alexiadis, V., Colyar, J. and Halkias, J. (2007) ‘A model endeavor’, Public Roads, Vol. 70, No. 4, Available: www. tfhrc.gov/pubrds/07jan/01.htm Gazis, D.C. (2002) Traffic Theory, Kluwer Academic Publishers, New York, Available: http://ebooks.kluweronline.com Herty, M., Klar, A. and Singh, A.K. (2007) ‘An ODE traffic network model’, Journal of Computational and Applied Mathematics, Vol. 203, pp.419–436. Phan, A. and Kien, C. (2007) ‘TP HCM ‘mat’ 14,000 ty dong vi ket xe’ (‘Ho Chi Minh City ‘lost’ VND 14 billion due to traffic jam’), Vietnam Express News on 21/9/2007, Available at: http://www.vnexpress.net/ Quadstone (2003a) Quadstone PARAMICS Version 4.1, Technical Notes Edinburgh. Quadstone (2003b) Quadstone PARAMICS Version 4.1, User Guides Edinburgh. 240 Q-L. Le Ngoc et al. Reynolds, C.W. (1987) ‘Flocks, herds, and schools: a distributed behavioral model’, Computer Graphics, Vol. 21, No. 4, pp.25–34. Takahashi, M., Nakanishi, T., Miyoshi, I. and Fujikura, T. (2002) ‘An evaluation of the road traffic system simulator PIMTRACS by PIM’, Mathematics and Computers in Simulation, Vol. 59, pp.45–56. General Statistics Office of Vietnam (2008) ‘Socio-economic, January 2008’, Monthly Statistical Information, available at: http://www.gso.gov.vn/default_en.aspx?tabid=501&thangtk= 01/2008 Wen, W. (2008) ‘A dynamic and automatic traffic light control expert system for solving the road congestion problem’, Expert Systems with Applications: An International Journal, Vol. 34, pp.2370–2381. Bibliography Cortes, P., Munuzuri, J., Ibanez, J.N. and Guadix, J. (2007) ‘Simulation of freight traffic in the Seville inland port’, Simulation Modelling Practice and Theory, Vol. 15, pp.256–271. Evans, J.R. and Olson, D.L. (2005) Introduction to Simulation and Risk Analysis, Pearson Custom Publishing, Boston, USA. Kara, S., Rugrungruang, F. and Kaebernick, H. (2007), ‘Simulation modelling of reverse logistics networks’, International Journal of Production Economics, Vol. 106, pp.61–69. Kelton, W.D., Sadowski, R.P. and Sturrock, D.T. (2004) Simulation with Arena, 3rd ed., McGraw-Hill Professional, New York. Law, A.M. (2007), Simulation Modeling and Analysis, 4th ed., Tata McGraw-Hill Publishing Company Limited, India. Oketch, T. and Carrick, M. (2005) Calibration and Validation of a Micro-Simulation Model in Network Analysis, Paper #05-1938, Presented at the TRB Annual Meeting, Washington DC. Taha, H.A. (2007) Operations Research: An Introduction, 8th ed., Prentice-Hall, New York. Wahle, J., Annen, O., Schuster, Ch., Neubert L. and Schreckenberg, M. (2001) ‘A dynamic route guidance system based on real traffic data’, European Journal of Operational Research, Vol. 131, pp.302–308.