Accurate traffic speed prediction is a critical precursor for traffic management activities such as congestion prevention, accident warning, and traffic network optimization. Given the complexity of traffic flow data and the spatiotemporal correlations within this data, we present a Spatiotemporal Multi-head Graph Attention (STMGAT) traffic speed prediction model, which consists of multiple layers for extracting spatiotemporal features from traffic speed data. Each layer is constructed by serially connecting a Bidirectional Long Short-Term Memory Network (Bi-LSTM), a Multi-head Graph Attention Network (MGAT), and a gated causal Convolution Neural Network (gated causal CNN). In this structure, Bi-LSTM and gated causal CNN are responsible for two stages of temporal feature extraction from the traffic speed, while the MGAT, positioned between Bi-LSTM and gated causal CNN, facilitates the transfer of temporal features and the aggregation of spatial features. Furthermore, a Bayesian Optimization-Evolutionary Algorithm (BO-EA) mutual feedback optimization framework is proposed to optimize the network architecture and hyperparameters of the STMGAT model. The evolutionary algorithm provides evolutionarily-informed sampling points to the Bayesian optimization, while the Bayesian optimization supplies exogenous population individuals to the evolutionary algorithm. The two algorithms effectively optimize the network architecture and hyperparameters of the STMGAT model by continuously exchanging optimization information with each other. Experimental results demonstrate that the STMGAT model, featured with a decision-based and adaptive architecture, can effectively capture the complex spatiotemporal features in traffic speed data, achieving superior prediction performance compared to the baseline models. Moreover, the BO-EA mutual feedback framework shows higher efficiency on optimizing the network architecture and hyperparameters compared to single optimization algorithms.

Short-term traffic speed prediction is fundamental to intelligent transportation systems (ITS), and the accuracy of the model largely determines the performance of real-time traffic control and management. In this study, a short-term traffic speed prediction method based on the spatial-temporal analysis of traffic flow and a combined deep-learning model, and a hybrid spatial-temporal feature selection algorithm (STFSA) of a convolutional neural network–gated recurrent unit (CNN-GRU)) is initially developed. Specifically, the STFSA is firstly employed to reconstruct the spatial-temporal matrix of traffic speed based on temporal continuity and spatial characteristics, and then this matrix is considered as the input feature of the prediction model. After this, the nonlinear fitting ability of the CNN is adopted to extract deep features from the convolutional and pooling layers for model training. Finally, by combining the timing and long-range dependence of the captured data with the forward GRU and the reverse GRU, the accuracy of the prediction result is further improved. The validity of the proposed model can be verified by comparing the prediction results with the actual traffic data. Accordingly, in the case study, the performance is compared with various benchmark methods under the same prediction scenario, verifying the superiority of the proposed model.

Predicting how long a trip will take may allow travelers plan ahead, save money, and avoid traffic congestion. The journey time estimation model should take into account three crucial factors: (1) individual travel preference, (2) dynamic spatio-temporal correlations, and (3) the association between long-term speed forecast and travel time estimate. In order to overcome these challenges, this study proposes a unique parallel architecture called the multi-task spatio-temporal attention network (MT-STAN) to estimate journey times. To extract the dynamic spatio-temporal correlations of the road network, we first develop a traffic speed prediction model based on spatio-temporal block and bridge transformer networks, combining the road, timestamp, and traffic speed information into hidden states. Second, we offer a personalized model for estimating journey times that makes use of cross-network, holistic attention, and semantic transformer. In this approach, travel preferences extraction through cross-network, holistic attention permits correlations between the dynamic road network’s hidden states and individual journey characteristics, which are subsequently transformed into global semantics by the semantic transformer; preferences and semantics are integrated during the estimate phase. Finally, a multi-task learning component is included, which combines both traffic speed prediction and individual journey time estimate, via the sharing of underlying network parameters and the improvement of the contextual semantic knowledge of the latter job. Evaluation experiments are carried out using a highway dataset collected in Yinchuan City, Ningxia Province, China. The proposed prediction model outperforms state-of-the-art baseline approaches in experiments.

The real-time performance and accuracy of traffic flow prediction directly affect the efficiency of traffic flow guidance systems, and traffic flow prediction is a hotspot in the field of intelligent transportation. To further improve the accuracy of short-term traffic flow prediction, a short-term traffic flow prediction model based on traffic flow time series analysis, and an improved long short-term memory network (LSTM) is proposed. First, perform time series analysis on traffic flow data and perform smoothing and standardization processing to obtain a stable time series as model input data, which can improve the accuracy of model training and eliminate the impact of a wide range of feature values. Then, an improved LSTM model based on LSTM and bidirectional LSTM networks are established. Combining the advantages of sequential data and the long-term dependence of forwarding LSTM and reverse LSTM, the bidirectional long-term memory network (BILSTM) is integrated into the prediction model. The first layer of the LSTM network learns and predicts the input time series and further learns and trains through the bidirectional LSTM network to effectively overcome the large prediction errors. Finally, the performance of the proposed method is evaluated by comparing the predicted results with actual traffic data. The model that is proposed in this paper is compared with the long short-term memory network (LSTM) model and the bidirectional long-term memory network (BILSTM) model. The results demonstrate that the proposed method outperforms both compared methods in terms of accuracy and stability.

Electric vehicles (EVs) endow great potentials for future transportation systems, while efficient charge scheduling strategies are crucial for improving profits and mass adoption of EVs. Two critical and open issues concerning EV charging are how to minimize the total charging cost (Objective 1) and how to minimize the peak load (Objective 2). Although extensive efforts have been made to model EV charging problems, little information is available about model properties and efficient algorithms for dynamic charging problems. This paper aims to fill these gaps. For Objective 1, we demonstrate that the greedy-choice property applies, which means that a globally optimal solution can be achieved by making locally optimal greedy choices, whereas it does not apply to Objective 2. We propose a non-myopic charging strategy accounting for future demands to achieve global optimality for Objective 2. The problem is addressed by a heuristic algorithm combining a multi-commodity network flow model with customized bisection search algorithm in a rolling horizon framework. To expedite the solution efficiency, we derive the upper bound and lower bound in the bisection search based on the relationship between charging volume and parking time. We also explore the impact of demand levels and peak arrival ratios on the system performance. Results show that with prediction, the peak load can converge to a globally optimal solution, and that an optimal look-ahead time exists beyond which any prediction is ineffective. The proposed algorithm outperforms the state-of-the-art algorithms, and is robust to the variations of demand and peak arrival ratios.

Various customized bus route optimization methods based on certain conditions have been applied to the actual route optimization problems, but the actual operation process of customized buses mostly lies in an uncertain condition. In this paper, a three-stage hybrid coding method based on NSGA-II algorithm was proposed to deal with customized bus route optimization under uncertain condition. Firstly, with the objective of minimizing passenger travel time and customized bus carbon emission, a robust optimization model was constructed. Second, with the Bertsimas-Sim robust optimization theory, the robust peer-to-peer transformation was performed on the robust model with uncertain parameters. Finally, the practical issue including three customized bus parking lots and 20 boarding and alighting stations were solved to verify the rationality of the model and algorithm. Compared with the hybrid algorithm based on K-means and multi-objective genetic algorithm, this method reveals not only better solution results, but saves 42.11% of computing time. The results are of great value for exploring customized bus route optimization methods and improving the efficiency of customized bus operations.

Understanding human driving style and classifying driver's risk pattern is the basis of traffic risk management. The recent rapid increase of the availability of taxi trajectory data, combined with the popular analysis techniques for big data, gives the chance of thorough analysis of taxi drivers' driving style and risk pattern. In this paper, the driving characteristics of 10674 taxies (at Qiangsheng Taxi Corporation) in a month are extracted from trajectory data. The trajectory data includes time, position, motion, as well as operating status. The method adopted in this paper is entropy weight-analytic hierarchy process (Entropy-AHP) with speed, over speed behavior, driving stability, mileage and time, and fatigue driving as first-grade indexes. The weights of indexes and risk value are calculated, then all taxi drivers are grouped into five risk grades. The risk pattern recognized from the data could be particularly helpful for insurance companies to formulate differentiated pricing strategy.