Balancing Transit Signal Priority: A Solution for Private Vehicle Delay

By Soham NandiReviewed by Susha Cheriyedath, M.Sc.Oct 18 2023

In an article published in the journal Nature, researchers propose a bi-level programming model that relies on the arrival rate and dynamic cycle of private vehicles in an attempt to improve transit signal priority systems so they do not negatively impact private vehicles by prioritizing public over private transport.

Background

Along with economic development, the rate of urbanization in China is also hugely increasing, leading to increasing problems in urban transportation, in other words, traffic. For public transportation, a very innovative solution was created called Transit Signal Priority (TSP), which has passive and active priority strategies.

TSP is more efficient than bus lanes; however, it focuses too much on public transportation and does not take into account private vehicles, in some cases causing delays to private vehicles just to create a more efficient path for the public ones. To address this, the authors of the present study propose a new framework named Dynamic Priority, which consists of three components: continuous detection, a signal control algorithm, and communication links.

Contributions of the Paper

The study made the following contributions:

• A bi-level programming is proposed to balance the benefits of the priority phase and the negative impacts of the non-priority phase
• A Game theory approach to achieve the dynamic cycle that helps in improving TSP
• A case study performed in the city of Zhengzhou, China, with a self-driving bus to present the effectiveness of the proposed method.

Most past studies on this topic were from the perspective of the buses; hence it is unreasonable to expect them to take into account the delay caused to private vehicles. This problem is tackled by the present study. The researchers used a self-driving bus as the subject. Due to the requirement of a connected vehicle environment, a few assumptions had to be made including (1) both the bus and the private vehicle data can be obtained and (2) using the rate of arrival and saturation flow rate the delay triangle can calculate private vehicle delay.

How Does the TSP System Work?

In this system, Vehicle-to-Everything (V2X) technology is used to gather real-time traffic data, including traffic status, traffic signals, other vehicles, infrastructures, and more. When the self-driving bus crosses the intersection, the roadside unit (RSU) detects it and private vehicles by establishing a connection with the on-board unit (OBU). The data from the self-driving bus and the private vehicle will be forwarded to the mobile edge computing server using a 5G network. Then, the final decision instructions will be sent to the Mobile Edge Computing server to the roadside unit, which then decides whether to implement the signal control strategy if there is a self-driving bus approaching the intersection, and then the instructions will be provided to the annunciator.

TSP has two phases: the Priority phase and the Non-Priority phase. While the priority phase has improved efficiency, balancing it alongside the negative impacts of the non-priority phase is the critical problem in optimizing TSP. To deal with the negative impact caused by the non-priority phase, researchers attempted to create models that help find the optimal signal timing to try and reduce the delay that occurs in the non-priority phase of TSP.

In this paper, the authors proposed an integrated index that is the weighted sum of all delays in both priority and non-priority phases; all of this is determined subjectively. This paper provides a two-level programming model, which operates in accordance with a connected vehicle environment to decide the optimal dynamic signal timing.

The Bi-Level Programming Model

The data from the self-driving bus can be obtained with the help of the RSU and the OBU. Using the data, the researchers calculated three parameters: the time required for the bus to travel from the current position to the stop line (t_b) and the remaining time of signal, i.e. the time for the signal to change to a different color, t_g for light that’s currently green and t_r for light that’s currently red.

When t_b > t_r (when the bus reaches the stop line after the light changes from red to green) or t_b<t_g (when the bus reaches the stop line before the light changes from green to red), the bus can easily cross the stop line, in that case, the original signal phase is implemented.

However, when t_b<t_r (when the bus reaches the stop line before the color changes from red to green) or t_b>t_g (when the bus reaches the stop line after the color changes from green to red), the bus cannot cross the stop line, according to original signal phase, so the dynamic cycle is inserted. Green Extension (extending the time the green light stays on) and Red Truncation (decreasing the time the red light stays on) are applied, which provides priority to the self-driving buses and prevents delay to the private vehicles.

In bi-level programming, the upper and lower levels are connected in a way that the upper model performs as the delay model whereas the lower model functions as the dynamic phase model, which means the lower model calculates the necessary variables based on the data gathered by RSU and OBU, and these data are passed on to the upper model to calculate the delay.

The Car-Following Model

Self-driving cars can use V2X to obtain the traffic status of the front vehicle and the surrounding traffic conditions. The PATH laboratory proposed the Cooperative Adaptive Cruise Control (CACC) model, which helps create a feedback control system that leads to automatic following of autonomous vehicles. This allows the speed of the following vehicle to be adjusted according to the preceding one, which allows the maintenance of safe distance between the two.

Genetic Algorithm: The Solution

Using a Genetic Algorithm, the authors attempted to find the optimal solution. As the name suggests, genetic algorithms work according to the concept of passing of genes to offspring and evolving. Here, each solution is like a person, with a unique code or solution, which is made up of 24 binary strings. This code represents how traffic lights should work to reduce delays in private vehicles.

There are 200 solutions (peoples) and these solutions evolve; the best or fittest solutions are chosen to be parents, and then the parents pass their code to their kids making even better solutions. This is repeated for 50 rounds to improve the solutions, after which the best solution is picked.

Conclusion

The TSP system is extremely crucial and efficient in the current road situations with heavy traffic and autonomous vehicles. However, its inability to balance the benefits of both buses and private vehicles is the focus of this study, more specifically how it causes delays to private vehicles to prioritize public transport.

The authors proposed a bi-level solution to effectively improve the TSP, taking into account the perspective of private vehicles and using a genetic algorithm to create the final model. This solution would help the urban traffic run smoothly and efficiently, considering more variables and subjective decision-making. It can be implemented in bus lines, isolated intersections, and multiple roads.

Based on case studies and simulations performed in the SUMO toolkit, the model provided by the genetic algorithm was shown to perform consistently, verifying the model's effectiveness. The delay per vehicle in this method was also found to be comparatively less than that in other methods.