Empirical Study on Maximum Traffic Throughputs at Intersections

Traffic congestion in big cities of Thailand has been a major problem seriously deterring economy growth. An average travel speed has become lower than 15 km/hour during peak-hour periods. One of the current major questions is how to find the most optimum green time that minimizes vehicle delay at an intersection. Since the identification of this optimum green time significantly depends on the knowledge of how fast vehicles can pass through an intersection, the information of the maximum traffic throughputs at intersections at different locations is indispensable. In this study, therefore, traffic counts were extracted manually from the videos recorded at 4 major intersections in Saensuk city. After the data were analyzed, the maximum traffic throughputs for different lane numbers and configurations were calculated and summarized. A better understanding of this maximum traffic throughput at intersections will lead to more accurate estimation of optimal green time.


Introduction
Traffic congestion in Thailand has been a major problem slowing down economy growth. With the number of vehicles exceeding the available roadway capacity in the network, traffic condition during rush hours becomes extremely congested. For example, at present, there are about 7 million registered vehicles in Bangkok but the current roadway capacity is only available for 2.5 million vehicles. The rush-hour periods expand longer each year mainly because of the increasing number of personal vehicles.
For last 20 years, to minimize traffic congestion, several new roads and expressways were built. However, there has been no sign of improvement on travel time in the area. Under bad economy for about a decade and the unsuccessful outcomes from previous traffic congestion mitigation strategies, most of major cities have limited budget and less support for new road construction projects from the communities. Therefore, an improvement on the current operations of existing traffic systems becomes more feasible option.
In 2015, after experiencing severe traffic congestion on major roads during peak hours, Saensuk city started a new traffic mitigation program. This new program's first phase started the work on major intersections with the highest congested level. The program emphasizes on the application of Intelligent Traffic Systems (ITS) to lessen the traffic congestion level. At the selected intersections, there was an attempt to use vehicle data collected from recorded videos to enhance an understanding of traffic congestion and identify the optimum green time that results in high intersection capacity. To be able to select the optimum green time, a good understanding of possible maximum traffic throughput at intersection is essential. As a result, the data on maximum traffic throughput for different lane numbers and configurations were collected and summarized in this study.

Literature review
There has been many researches with the effort to minimize vehicle delay at intersections. The work is related to important traffic parameters such as traffic capacity and saturation flow rate at an intersection. Samples of the researches can include the following works. In 1968, Miller [1] conducted a research to estimate capacity at signalized intersections in Australia. After that, Akcelik [2] analyzed not only capacity but also timing at signal intersections. Kimber et al. [3] developed a procedure to predict the saturation flow rate at an intersection that is controlled by traffic signals. Bester and Meyers [4] also worked on saturation flow rate at an intersection.
In recent years, non-motorized mode has become more popular. As a result, Li et al. [5] conducted a research on the saturation flow rate of shared nonmotorized lane at intersections. Applying a better understanding of the traffic throughput at an intersection, many researchers had developed efficient ways to improve traffic throughputs, such as Liu and Chang who worked on an arterial signal optimization for intersections [6] and Li and Elefteriadou who worked on maximizing the throughput of turn bays at a signalized intersection [7]. Additional works may include the work from Zhou et al. [8], Wu et al. [9], and Chunhui et al. [10]. As advanced traffic-related and automobile technologies are rapidly growing and increasingly play an important role in our present society, new research related to the reduction of vehicle delay at intersections has become broader.

Data source
Intersections with high traffic congestion level in the Saensuk city were selected for this study, which include 4 major intersections that are 1) Galaxy, 2) Burapa, 3) Kao Larm, and 4) Lam Tann. These intersections are on the major arterials in the city. Each intersection has different number of arms and lanes and/or lane configurations. For example, the Galaxy intersection is a three-way junction as shown in Fig. 1. The Southbound (SB) approach has 4 lanes. Similarly, the Northbound (NB) approach has 4 lanes. However, the Eastbound (EB) approach has 3 lanes. The lane type is also displayed by white arrows in Fig. 1. The Burapa intersection is also a three-way junction with 2 lanes EB, 3 lanes WB, and 3 lanes NB (shown in Fig. 2). The special thing at this intersection is the rightmost u-turn lane of the WB approach. The last two intersections are Kao Larm and Lam Tann intersections, which are four-way junctions ( Fig. 3 and 4, respectively). The Kao Larm intersection has 2 lanes EB, 4 lanes WB, 2 lanes NB, and 3 lanes SB. Similarly, for the Lam Tann intersection, there are 1 lane EB, 2 lanes WB, 2 lanes NB, and 3 lanes SB.
To collect traffic count data, the CCTV cameras were installed at these 4 major intersections (displayed in Fig. 2). Videos were recorded for 24 hours for the days with high traffic congestion. After the investigation of traffic condition, the time interval 16:00-20:00 has been identified as the time period with highest traffic volume. Therefore, the manual traffic count was conducted for this time period and the results are summarized in this study.

Maximum traffic throughout
To ensure the accuracy of the collected traffic data, vehicle counts were extracted manually for three signal cycles in every half hour between 16:00-20:00 and shown in Tables 1-4  At the Burapa intersection, since Lane 1 is for through and U-turn traffic, its maximum traffic throughput is less than that of Lane 2, which is throughleft-turn lane. At this intersection, traffic demands on Lane 3 NB, Lane 2 EB, and Lane 3 WB are significantly less than the capacity, so those lanes were not included in Table 2.     Table 5.
The surprising finding is that the traffic throughputs on Type 1 lane is never higher than 1,500 vehicles/hour and the traffic throughputs for Types 2 and 3 lanes are always higher than that of Type 1 lane. According to Table 5, the medians of vehicle numbers passing through the selected intersection on Types 2 and 3 lanes are 1,310 and 1,402 vehicles/hour, but Type 1 lane has only 1,088 vehicles/hour.

Conclusions and recommendations
In this study, the information of possible maximum traffic throughputs at major intersections was collected using the traffic count data extracted from the videos recorded by deploying CCTV cameras. Four major intersections with high traffic congestion in Saensuk city were chosen and afterward new CCTV cameras were installed at these intersections. Traffic throughput data were manually extracted from the recorded videos for the time period of 16:00-20:00 because, after the visual observation, this time period covers peak traffic hours. The maximum traffic throughputs for different number of lanes and lane configurations at the chosen intersections were collected and analyzed. There are two 3-way and two 4-way intersections. Traffic throughputs for different lane types should be treated separately. With a better understanding of possible maximum traffic throughputs, the identification of optimum green time to minimize vehicle delay would become more accurate. In the future work, a statistical analysis to identify a relationship among important parameters, such as green time vs. maximum throughput will be conducted.