Nowadays, congestion is becoming an event that is common to see every day. As the people’s mobility increase, the number of vehicles that roam the road also increase. This increase in number of vehicles sometimes leads to severe congestion, if the traffic flow wasn’t managed properly. Severe congestion usually happened if vehicles input flow exceeding the road capacity, and leading to the gridlock condition (vehicles from one link is blocked by the vehicle’s spillback form the other link in an intersection). If the gridlock condition occurs, it will be needed long time to recover the traffic condition.

In order to prevent this gridlock condition, traffic control scheme is needed to be applied. There are many kind of traffic control algorithm that has been implemented in many cities around the world such as SCATS, SCOOT, TUC, or OPAC PRODYN. Most of those algorithms work by calculating the most effective green time allocation for each link, by taking into account the vehicles flow, link density, and many other traffic parameters. Each algorithm has its own advantages and disadvantages, therefore it’s difficult to conclude which algorithm is the best. In the last 10 years, a new traffic algorithm based on distributed control scheme has been developed in hopes of solving the gridlock problem.

This algorithm is an adaptation of a communication network control algorithm, which studied in 1992. Simply put, this algorithm works by comparing the traffic demands between upstream and downstream link to get weighting value of each link. This way, the algorithm will prevent the upstream vehicles overflow the downstream link, because each calculation always take into account the queue vehicle at the downstream link.

One thing that differentiate this algorithm and the other traffic control algorithms is that this algorithm works in a distributed control scheme, not a centralized control scheme, where the control is done ‘locally’ at the respective intersection (illustrated in figure 2). Using the algorithm in a centralized manner implies the need for concentration of all queue information and rather complex constraint sets and computations, especially for larger networks. Through distributed control scheme, the control architecture becomes much simpler, since the control scheme is limited in each intersection, with little communication between intersections. In addition, because of its simple algorithm, it does not require exceptionally complex hardware capabilities. Another advantage of this distributed control concept is that if there is a control failure at an intersection, other intersections will not be affected by this failure.

In order to implement the proposed algorithm in Bandung area, PRE has to do it in the simulator environment. This is because the government of Bandung has restricted the access to the existing traffic light infrastructure to all parties other than Department of Transportation. This restriction is set by the government to prevent the traffic chaos if the trial of new algorithm leads to the failure state. Therefore, the only way to implement the algorithm is by using traffic simulation software. Moreover, in order to simulate the actual Bandung traffic condition, one needs to build the network model that could represent the actual traffic condition. Therefore, we have done some preliminary researches to get actual network data such as traffic flow, green time allocation, and turning probability in every intersection. That way, we could simulate the actual traffic condition in the simulator environment, and implement the algorithm in that model. Figure 3 shows the network area that was chosen as the area to implement the proposed algorithm.

In this simulation we compared the traffic condition controlled using conventional Fixed Time (FT) control, and two proposed algorithms (BPS & BPC). Through the simulation done for the modelled network, we found out that in actual condition, congestion in intersection 2 occurs as a result from very high traffic flow from Lembong Rd that moves to Tamblong Rd and Veteran Rd (~9000 veh/hr). As shown in figure 6.a, FT control couldn’t handle the enormous traffic flow well, resulting to a very long queue in Tamblong Rd. Evaluation shown in figure 7 shows that average queue length during this peak period reaches 117 m using traditional FT control. However, when we applied BPS and BPC, the results is very different. The average queue length using BPS control only reaches 27 m, while BPC control only reaches 49 m. It means, the algorithm could reduce the queue length in Tamblong Rd (Intersection 2) up to 77% using BPS control, and up to 67% using BPC control, compared to FT control.