Real-time traffic jams prediction inspired by Biham, Middleton and Levine (BML) model

Highlights

• A simple and efficient M-BML model is proposed for urban traffic.

• An effective way is presented to map real routes into the M-BML model.

• A novel strategy is constructed to map cells of the M-BML model into real intersections.

• The M-BML model can predict the traffic jams at real intersections in real time.

Abstract

Urban traffic jams severely affect many cities around the world. Therefore, there is an urgent need for accurate prediction of traffic jams in real time. With many existing approaches, it is difficult to obtain favorable results for dynamic and uncertain traffic flows in urban networks. This paper proposes a real-time traffic jam prediction method based on a two-dimension cellular automaton model inspired by the well-known Biham, Middleton and Levine (BML) model. The proposed model combines the simplicity and high efficiency of the BML model with real urban traffic networks. Our work can be divided into two parts: (1) the proposal of a practical approach for mapping the urban traffic topological structure into a modified BML (M-BML) model, and (2) the development of a novel strategy to deal with different cells of the M-BML model, especially conflict points and fuzzy points, which facilitates the recognition of real traffic jams at intersections. To the best of our knowledge, this is the first time that real urban traffic road networks are directly mapped into a descriptive model for jam prediction. Extensive experiments confirm the accuracy and efficiency of the proposed M-BML model in predicting traffic jams in dynamic and uncertain real urban traffic networks.

Introduction

Traffic jams are among the most influential challenges in today's society, especially in metropolises. They act as major obstacles to many industries by reducing their efficiency, undermining their reputation, and causing great economic loss compounded by environmental damage [14], [15]. Moreover, individuals suffer from severe delays due to traffic jams in daily life. Therefore, alleviation of traffic jams has emerged as a topic of considerable interest in the field of intelligent transportation. Increasing the physical system capacity through construction is usually infeasible; hence, the development of intelligent transportation systems (ITS) is a promising approach [9], [31] for achieving real-time traffic control and dynamic traffic management, which can alleviate traffic jams in metropolises and enhance transportation efficiency. Short-term traffic flow forecasting is a fundamental task for many ITS subsystems such as advanced traveler information systems (ATIS), advanced traffic management systems (ATMS) [26]. If traffic jams at a location could be predicted in advance, it would be much easier for the traffic management center to react and implement efficient solutions, such as changing the parameters of the traffic lights nearby, using variable message signs, or sending information to drivers by means of radio broadcast. In other words, accurate and real-time short-term prediction of traffic jams is a key element of ITS operation optimization. Owing to the insufficiency of existing traffic jam prediction methods, this paper explores an effective method for accurate and real-time traffic jams prediction in urban networks.

Current models for short-term traffic forecasting mainly involve two types of approaches. In the early years, most models adopted classical statistical approaches for predicting traffic at a single point, such as the autoregressive integrated moving average (ARIMA) model [4], Kalman filtering model [32], and multivariate adaptive regression splines (MARS) [35]. These models provide perfect solutions to most time-series problems [30]. However, when faced with extensive datasets comprising structured and unstructured data, most classical approaches have been shown to be weak or inadequate, especially under unstable traffic conditions and complex road settings. With the recent proliferation of computational intelligence (CI) approaches for data analysis, researchers have been focusing on neural and evolutionary computational models [17], [21], [22], [36]. The most widely used CI models are artificial neural networks (ANN) [13], including back propagation neural networks (BPNN) and some hybrid ANN models such as the Bayesian combined neural network model as well as models involving fuzzy and evolutionary techniques. The ANN model is suitable for arbitrary functions, especially nonlinear functions, but it suffers from a disadvantage in the case of non-convex problems, i.e., difficulty in understanding the objective function and finding the global optimal solution [1]. Support vector machines (SVM) can effectively overcome the shortcomings of ANN. They can not only use the minimal risk strategy for training, but also use structural risk minimization strategies for minimizing the upper bounds of errors. SVM can guarantee the global optimal value in theory, whereas ANN can only obtain the local optimal value. However, both fail to consider the conditions when the traffic data include noise [18]. Despite the obvious progress in the above-mentioned methods, they usually predict the traffic flow of each road separately (e.g., the ARIMA model) or use too many assumptions and constraints; hence, they may be limited and less effective in practical situations. Moreover, they performed the traffic prediction under determinative and known circumstances; hence, they cannot describe the intrinsic randomness or uncertainty in the urban traffic management problem effectively [16], [24].

To effectively describe traffic flows under a variety of circumstances, researchers proposed an uncertainty-based cellular automaton model, and later, a binary dimension cellular automaton model known as the Biham, Middleton and Levine (BML) model [3]. The BML model is a promising tool for modeling urban traffic networks. It can not only describe the different states of urban traffic networks but also identify the key factors affecting phase transitions, such as traffic density and traffic assignment. To represent urban transportation more practically, scholars have proposed many extensions of the basic BML model. Nagatani [27] proposed an improved BML model by introducing the cloverleaf junction in order to reduce gridlocks in the BML and improve the running status of traffic streams. Fukui [12] considered the average speed in the BML and used the running speed of the traffic by introducing the individual high-speed vehicles. Chung [6] analyzed the effect of broken traffic lights on transportation. It was revealed that an increase in the number of broken traffic lights increases the probability of not only traffic jams at intersections but also gridlocks, while the resulting variable density decreases. Cuesta [7] and Nagatani [27] were the first to introduce switch rules into the BML model. Freund [11] investigated the binary direction transportation problem. Other researchers have explored the mean field theory in the BML model [8].

Although current BML models can describe most features of traffic flows, they cannot predict real traffic jams because they lack information on the topological structure of road networks, regardless of their ability to accurately determine sites where traffic jams occur [23], [29]. It has been shown that intersections are the main sites of traffic jams. Accurate and real-time traffic jams prediction for intersections can support real-time route guidance and reliable traffic control strategies, thereby alleviating traffic jams. However, most current methods, including all the BML-based methods, might not be able to predict traffic jams at real intersections effectively [19], [20], [28], [34], because only a few of them take intersections into consideration [33]. Thus, accurate real-time prediction of urban traffic jams at intersections is a challenging task.

This paper proposes a modified BML (M-BML) model that combines the simplicity and high efficiency of the BML model with real urban traffic networks. The objective of the proposed model is to accurately predict urban traffic jams in real time. The proposed model mainly differs from state-of-the-art methods by focusing on real-time traffic jam prediction at intersections. All the factors related to intersections are taken into consideration, and the prediction results are time-mapped to actual jammed intersections in real traffic situations.

The contributions of this paper can be summarized as follows.

• An effective approach is proposed for mapping the real urban traffic network topological structure into the M-BML model. Real routes in urban traffic networks are abstracted and mapped into the lattice of the BML model, and their traffic densities are obtained for initialization of the model.

• A novel strategy is developed to deal with different cells of the M-BML model, especially conflict points and fuzzy points, which facilitates the recognition of real traffic jams at intersections. Based on the high efficiency of the M-BML model and the novel mapping strategy, real traffic jams in urban traffic networks can be predicted within 15 minutes with high accuracy and efficiency.

The remainder of this paper is organized as follows. Section 2 describes the BML model, based on which a modified BML (M-BML) model is proposed. Section 3 details the bidirectional mapping between an urban traffic network and the M-BML model. Section 4 discusses the prediction of traffic jams on the basis of the M-BML model. Section 5 presents the experiment settings and results. Finally, Section 6 concludes the paper and briefly explores a direction for future work.