Multimodal transport hubs are pivotal nodes for improving urban transportation efficiency, and accurate passenger flow prediction constitutes the core of their operational management. Although massive volumes of data can be collected via Internet of Things (IoT) sensing systems, data missing frequently occurs due to equipment malfunctions and communication interruptions. Coupled with the inherently complex spatio-temporal dependencies of passenger flow, traditional methods prove ineffective in such scenarios. Existing studies mostly adopt the paradigm of imputation first, then prediction, which easily induces error propagation and lacks robustness under incomplete data conditions. To address this challenge, a novel Structured Decomposition and Integration (SDI) framework is proposed for robust passenger flow prediction with incomplete data. First, a Spectral-Guided Hierarchical Variational Mode Decomposition (SGH-VMD) method is designed to decouple the aliased modes of complex passenger flow sequences, yielding highly representative decomposed features. Second, with these decomposed features as inputs, a Dual Generative Adversarial Network with Gradient Penalty (Duel-GAN-GP) is constructed to collaboratively complete missing value generation and future sequence prediction, and aggregate the prediction results to reconstruct an integrated passenger flow structure. In addition, the model integrates an efficient self-attention mechanism with multi-scale positional encoding to capture long-range temporal dependencies and model the complex evolution process of passenger flow. Comprehensive validations are conducted on a real dataset from a large-scale multimodal transport hub. Experimental results show that the SDI framework outperforms mainstream baseline models in both accuracy and stability across various artificially simulated incomplete data scenarios, and can fully meet the requirements of real-time applications. Ablation experiments further verify the independent contributions of each core component, providing an effective theoretical solution for modeling complex spatio-temporal passenger flow under incomplete data conditions.

Vehicular ad hoc networks (VANETs) hold significant potential for enhancing road safety and traffic efficiency. The performance of VANETs heavily relies on the strategic deployment of roadside units (RSUs), which gather and disseminate critical information. A key challenge is that different types of information possess varying timeliness requirements, rendering delayed information ineffective. However, this crucial aspect has not been sufficiently addressed in the existing RSU deployment literature. To bridge this gap, we first classify operational scenarios based on discretized traffic flows and information types, analyzing their corresponding transmission time constraints. We then formulate an RSU deployment model that explicitly incorporates these heterogeneous timeliness requirements. To solve this complex problem, we develop a Branch-and-Benders decomposition (BBD) algorithm, which partitions the problem into a master problem for determining RSU locations and multiple subproblems for allocating vehicle communication demands in each scenario. The master problem is solved using a branch-and-cut procedure. Upon finding an integer feasible solution, the dual subproblems are solved to generate optimality and feasibility cuts that are dynamically added to the master problem. Furthermore, we introduce valid inequalities to accelerate convergence. Numerical experiments demonstrate that the proposed BBD algorithm can efficiently generate provably high-quality solutions.

Efficient identification of key nodes is crucial to optimizing detector deployment and enhancing traffic monitoring in intelligent transportation systems. However, existing approaches often struggle to adapt to dynamic traffic variations, leading to suboptimal coverage and increased deployment costs. We propose bidirectional temporal convolutional graph attention networks (BTC-GATs) to address these limitations. This novel framework integrates bidirectional attention mechanisms to capture upstream and downstream dependencies, temporal convolutional networks for multiscale feature extraction, and graph attention networks for spatial information aggregation. BTC-GATs incorporates adaptive temporal modeling to capture nonlinear traffic dynamics, gradient-based variation analysis to quantify node influence, and a ranking mechanism that fuses attention coefficients with topological attributes to further enhance robustness and interoperability. In addition, a key node coverage study is conducted to examine the trade-off between accuracy and deployment efficiency. Extensive experiments on the California Highway PeMS04 dataset demonstrate that BTC-GATs outperforms benchmark methods in key node identification, offering superior accuracy and stability. Further analysis confirms its robustness under varying traffic conditions and initialization settings, highlighting its potential as a scalable, adaptive, and cost-effective solution for intelligent traffic monitoring. By facilitating efficient sensor placement, BTC-GATs contributes to improved data collection and congestion management in large-scale transportation networks.

Vehicular Ad hoc Networks (VANETs) in intelligent transportation systems have been regarded as an effective means to alleviate traffic congestion, reduce traffic accidents and save fuel. A proper roadside unit (RSU) deployment strategy is essential to improve the efficiency and stability of VANETs’ communication. However, the RSU deployment strategy that combines coverage and quality of service is still limited. To provide city planners with decision-making support, a multi-objective optimization model is built to optimize the deployment of RSUs under a limited budget. Two objective functions are proposed to maximize the number of communication tasks served and minimize the total task-weighted delay. Through a simple transformation of the delay matrix, we transform the proposed multi-objective model into a p-median problem. Then, we design a Lagrangian relaxation algorithm in a multi-objective framework to solve the model. Moreover, simulation examples are presented to demonstrate and validate the proposed model and solution algorithm. We analyze the gap between the algorithm result and the optimal solution, and characterize the Pareto front. The simulation results verify the feasibility and effectiveness of the proposed model and algorithm.

The roadside unit (RSU), which enables vehicle-to-infrastructure communication, is essential for improving the communication performance of vehicular ad hoc networks. However, optimizing the deployment of RSUs while considering their deployment at both fixed locations and on mobile vehicles remains a challenging issue. To bridge this gap, we develop two spatio-temporal networks derived from vehicle trajectories. Subsequently, the joint deployment challenge of static and mobile RSUs is articulated as a mixed-integer programming model. After linearization, the model can be directly solved by CPLEX. Additionally, the integrated optimization problem can also be formulated as a route-based model. Due to the exponential growth of route numbers, a branch-and-price (BAP) algorithm is designed to solve the route-based model. Within the framework of the BAP, we develop a heuristic technique for generating effective initial solutions. Based on the characteristics of the spatio-temporal network, a directed acyclic graph shortest path algorithm is utilized to accelerate the solution of column generation pricing problem at each node. Simulation examples are presented to demonstrate the proposed algorithm. The results indicate that the BAP can generate verifiable high-quality solutions and has a significant speed advantage over CPLEX for large-scale problems. Furthermore, a series of sensitivity analyses are conducted to assess the system’s responses to various influencing factors.

Aiming at improving the operation of the bottleneck area of the highway in the environment of a connected and automated vehicle, this paper proposes an integrated lane-level control (ILC) method by combining the variable speed limit control method and lane selection method into a comprehensive framework. A lane-level variable speed limit (LVSL) control method is proposed for the mixed traffic flow based on a deep deterministic policy gradient algorithm. Then, a lane-level short-term traffic prediction (LSTP) model based on hybrid deep learning is built to forecast the traffic state in a next control horizon. Finally, a lane selection method using a dynamic programming algorithm is established for the connected automated vehicle (CAV) to look for the optimal lane-level route by considering the estimated traffic speeds and limit speeds from LSTP and LVSL respectively. Comprehensive simulation-based evaluation experiments were conducted in various scenarios e.g., with different traffic demands, penetration rates of CAVs, length of control horizons, and the number of lanes. The evaluation results reveal that the proposed LSTP model outperforms the state-of-the-art traffic prediction models in terms of accuracy and stability. In addition, the proposed ILC method is capable of improving the traffic operation of the bottleneck efficiently by synthetically taking the advantage of the LVSL method and lane selection method. Compared with the no-control strategies, the ILC method can reduce total travel time by more than 30% in various traffic scenarios.

Lane-level traffic state prediction is one of the most essential issues in the connected automated vehicle highway systems. Accurate and timely traffic state prediction of the lane sections can assist the connected automated vehicles in planning the optimal route and making lane selection. In this article, we tackle the problem of forecasting lane-level short-term traffic speed and propose a novel mixed deep learning (MDL) model by coordinating the convolutional long short-term memory (Conv-LSTM) layers, convolutional layers, and a dense layer in an end-to-end structure. The introduction of the Conv-LSTM neural network enables the proposed MDL model to better capture the spatio-temporal characteristics and correlations of the dynamic lane-based traffic flow synchronously. To improve the efficiency of the proposed model, a feature correlation analysis method based on the maximum information coefficient is presented to measure the relevance between the historical traffic flows and the traffic speeds to be forecasted. Validated by the ground-truth traffic flow data collected by the remote traffic microwave sensors installed on the expressways in Beijing, the MDL model is capable of capturing the fluctuation of the lane-level traffic speeds at different types of lanes effectively during the whole day. Furthermore, the results confirm that the MDL model achieves better predictive performance than several state-of-the-art benchmark models in terms of prediction accuracy and space-time distributions. Our code and data are available athttps://github.com/lwqs93/MDL.

Short-term traffic volume prediction, which can assist road users in choosing appropriate routes and reducing travel time cost, is a significant topic of intelligent transportation system. To overcome the error magnification phenomena of traditional combination methods and to improve prediction performance, this paper proposes an improved Bayesian combination model with deep learning (IBCM-DL) for traffic flow prediction. First, an IBCM framework is established based on the new BCM framework proposed by Wang. Then, correlation analysis is used to analyze the relevance between the historical traffic flow and the traffic flow within the current interval. Three sub-predictors including the gated recurrent unit neural network (GRUNN), autoregressive integrated moving average (ARIMA), and radial basis function neural network (RBFNN) are incorporated into the IBCM framework to take advantage of each method. The real-world traffic volume data captured by microwave sensors located on the expressways of Beijing was used to validate the proposed model in multiple scenarios. The overall results illustrate that the IBCM-DL model outperforms the other state-of-the-art methods in terms of accuracy and stability.