Cooperative control of connected automated vehicles (CAVs) offers a promising solution for reducing traffic congestion and accidents. However, existing optimization-based and search-based methods for trajectory planning and vehicle scheduling struggle with real-time multi-vehicle control. This paper introduces a hybrid bi-level approach that nests optimization modelling within deep reinforcement learning (DRL) to jointly optimize vehicle sequences, lane selections, and trajectories, providing a rapid, safe, and high-quality solution to enhance traffic performance at multi-lane freeway merging sections. Specifically, we approach the problem of lane selection and vehicle sequencing for multiple vehicles as a multi-step decision-making process. At the upper level, we design a DRL agent with an attention-based encoder-decoder structure that auto-regressively constructs driving sequences and lane choices. It generates a probability matrix to select the next passing vehicle and target lane based on prior decisions at each step. The attention mechanism enables the centralized upper level to adapt to scenarios with varying vehicle counts without the need to retrain. At the lower level, we formulate a model predictive control (MPC) planner to generate safe trajectories. The resulting travel delay guides the upper-level DRL agent learning to maximize overall traffic efficiency. Moreover, we introduce a leader-and-lane-specific credit assignment mechanism that leverages domain knowledge to link each action with associated travel delays. This mechanism enables the agent to accurately recognize the impact of decisions on total delay, enhancing learning performance. Simulation results suggest that the proposed approach’s superior real-time performance and scalability from several to over a dozen vehicles, making it well-suited for practical automated merging tasks.

With the rapid development of Large Language Models (LLMs), aligning these models with human preferences and values is critical to ensuring ethical and safe applications. However, existing alignment techniques such as RLHF or DPO often require direct fine-tuning on LLMs with billions of parameters, resulting in substantial computational costs and inefficiencies. To address this, we propose Micro token-level Accept-Reject Aligning (MARA) approach designed to operate independently of the language models. MARA simplifies the alignment process by decomposing sentence-level preference learning into token-level binary classification, where a compact three-layer fully-connected network determines whether candidate tokens are “Accepted” or “Rejected” as part of the response. Extensive experiments across seven different LLMs and three open-source datasets show that MARA achieves significant improvements in alignment performance while reducing computational costs. The source code and implementation details are publicly available at https://github.com/IAAR-Shanghai/MARA, and the trained models are released at https://huggingface.co/IAAR-Shanghai/MARA_AGENTS.

Surrogate-assisted evolutionary algorithms (SAEAs) are well-suited for optimizing computationally expensive black-box problems in diverse real-world scenarios. The sample efficiency of SAEAs depends largely on the configuration of the surrogate model and sampling criteria. However, configuring these core components requires substantial manual effort and expert knowledge, limiting the broader applicability of SAEAs. To address these challenges, we propose CoE-SAEA, a novel paradigm that co-evolves large language models (LLMs) and configuration strategies to enhance SAEAs. Specifically, the paradigm consists of three populations with distinct roles: one evolves LLM prompts to generate robust configuration strategy instructions, another optimizes the configuration strategies, and the third solves the optimization problem using the selected algorithm configuration. Additionally, an exploration-exploitation module is incorporated to decide whether to explore new configuration strategies via LLMs or exploit existing ones. We empirically validate the efficacy of CoE-SAEA by comparing it to state-of-the-art algorithms across various benchmark problems and a real-world traffic signal optimization task. The source code of the proposed CoE-SAEA is publicly available at: https://github.com/ForrestXie9/CoE-SAEA.

In urban environments, road closures due to construction, maintenance, accidents, or emergency situations pose significant challenges to traffic flow and safety. The cooperative lane-changing (CLC) is a promising solution. Traditional rule-based CLC models often fall short in addressing the complexities introduced by sudden lane reductions and diversions. Therefore, this paper proposes a vehicle-to-everything (V2X)-aided multi-agent CLC (MA-CLC) model, which is tailor-made for road closure scenarios. By leveraging V2X, the connected and autonomous vehicle (CAV) and human driven vehicle (HDV) agents can share information about speeds, distances, and trajectories, which will be used as part of the state space for training. To make the HDV more humanlike, we customize a human-imitating reward function for HDV agents and implement CLC experiments with human expert drivers (HEDs). The results show that the safety and efficiency performance of the proposed MA-CLC model is respectively 15% and 26% higher than other benchmarks on average.

Planning safe and efficient trajectories through signal-free intersections presents significant challenges for autonomous vehicles (AVs), particularly in dynamic, multi-task environments with unpredictable interactions and an increased possibility of conflicts. This study aims to address these challenges by developing a unified, robust, adaptive framework to ensure safety and efficiency across three distinct intersection movements: left-turn, right-turn, and straight-ahead. Existing methods often struggle to reliably ensure safety and effectively learn multi-task behaviors from demonstrations in such environments. This study proposes a safety-critical planning method that integrates Dynamic High-Order Control Barrier Functions (DHOCBF) with a diffusion-based model, called Dynamic Safety-Critical Diffuser (DSC-Diffuser), offering a robust solution for adaptive, safe, and multi-task driving in signal-free intersections. The DSC-Diffuser leverages task-guided planning to enhance efficiency, allowing the simultaneous learning of multiple driving tasks from real-world expert demonstrations. Moreover, the incorporation of goal-oriented constraints significantly reduces displacement errors, ensuring precise trajectory execution. To further ensure driving safety in dynamic environments, the proposed DHOCBF framework dynamically adjusts to account for the movements of surrounding vehicles, offering enhanced adaptability and reduce the conservatism compared to traditional control barrier functions. Validity evaluations of DHOCBF, conducted through numerical simulations, demonstrate its robustness in adapting to variations in obstacle velocities, sizes, uncertainties, and locations, effectively maintaining driving safety across a wide range of complex and uncertain scenarios. Comprehensive performance evaluations demonstrate that DSC-Diffuser generates realistic, stable, and generalizable policies, providing flexibility and reliable safety assurance in complex multi-task driving scenarios.

Headway regularity is an essential indicator of transit reliability, directly influencing passenger waiting time and transit service quality. In this paper, we employ multi-agent reinforcement learning (MARL) to develop a Cooperative Transit signal priority strategy with Variable phase for Headway adherence (CTVH) under a multi-intersection network. Each signalized intersection is controlled by an RL agent, which determines the next step’s signal, adapting to real-time traffic dynamics of transits and non-transits and promoting transit headway adherence. The proposed approach considers four critical aspects, i.e., complicated states with multiple conflicting bus requests, rational actions constrained by domain knowledge, comprehensive rewards balancing buses and cars, and a collaborative training scheme among agents. They are correspondingly addressed by proper state representation with estimated bus headway deviations, irrational actions masking, reward functions formulated by general traffic queue and transit headway deviation, and appropriate MARL approach with synchronous action processing. Our method also takes into account the phase transition loss by setting yellow and all-red time. Simulation results compared with the coordinated fixed-time signal (CFT) and bus holding (BH) strategy verify the merits of the proposed method in terms of improvements in transit headway adherence and influence on general traffic. Based on the results, we further discuss the BH method’s limitations due to bus bay length and various holding lines and the CTVH method’s benefits in the three-intersection environment and the entire-line network. The proposed method has a promising application in practice to improve transit reliability.

Accurate and real-time acquisition of vehicular system dynamic states, road surface conditions, and motion states of surrounding participants is crucial for the safety, passenger comfort, and operational efficiency of autonomous vehicles (AVs) and connected automated vehicles (CAVs). In recent years, a significant amount of research has contributed to the field of state estimation for vehicles, roads, and pedestrians. From the systemwide perspective of intelligent transportation systems to a focused view on “vehicle-road-pedestrian”, this survey aims to provide a comprehensive review and summary of recent state estimation techniques for vehicle motion, road surface, and pedestrian motion. A thorough analysis of the reviewed literature, relevant datasets, evaluation metrics, and experimental platforms in this field is also conducted. Finally, existing challenges and future research directions about methods and performance evaluation are further discussed. This survey is expected to contribute to the advancement of research in dynamic state estimation of vehicle-road-pedestrian, thereby facilitating the development of efficient and safe intelligent transportation systems.

The estimation of bus passenger origin-destination (OD) flow is crucial for bus operation and management. Although many existing studies have made efforts to reconstruct OD flow from passenger counts at each station, few of them tracked individual passengers because it is still infeasible in practice to monitor all onboard passengers throughout their entire journeys. This study proposes a passenger reidentification method for estimating route-level bus passenger OD flow using video images at bus doors. All boarding and alighting passengers are identified and their appearance features are extracted from video images. Boarding passengers at upstream stations and alighting passengers at downstream stations are matched considering not only the similarities of their appearance features but also historical alighting probabilities. Reliable passenger matches are filtered out by a matching probability threshold. The corresponding accurate OD samples, together with the boarding and alighting counts at each station, are utilized to estimate OD flow between bus stations. A comprehensive case study was carried out on an operational bus route in Shanghai, China during morning and evening peak hours over 15 weekdays. Results of the case study showed that the proposed method is capable of generating accurate OD estimates, with cosine similarities remaining above 0.9, root mean square errors within 0.5 person, and mean absolute percentage errors of approximately 12%. The error metrics were decreased by nearly half compared to the traditional method that only uses the boarding and alighting counts.

AutopilotingConnected and Autonomous Vehicles(CAVs) is an important application for mobile computing. A promising context to realize autopiloting CAVs is cooperative driving on dedicated highways. For such a context, an indispensable driving scenario isCooperative Lane Change(CLC). Due to the safety concerns of this driving scenario, a verifiably safe solution is needed (at least, the solution design should be formally provably safe). However, this demand is complicated by the inherently unreliable wireless communications between the CAVs. In this paper, we focus on the well-adoptedConstant Time Headway(CTH) safety rule. We propose a CLC protocol, and formally prove its guarantee of the CTH safety and liveness, even under arbitrary wireless packet losses. These theoretical claims are further confirmed by our simulations. The simulation results also show that our proposed protocol significantly improves lane change success rates (by$5.3\% \sim +\infty \%$) than other alternatives under adverse conditions. Furthermore, the sensitivity study results also show our protocol can tolerate reasonable disturbances.

Intensive interactions among vehicles at freeway on-ramp merging areas lead to congestion and accidents. The emergence of connected automated vehicles (CAVs) has shown great potential to improve this issue. In this paper, a mixed integer nonlinear programming (MINLP) model is proposed and solved for the task of a cooperative merging of two traffic streams at a freeway on-ramp merging section. The proposed model simultaneously optimizes multiple vehicles’ trajectories and their merging sequence to improve traffic efficiency and ensure safety. Unlike conventional treatments, which match one mainline facilitating vehicle with one merging vehicle, the proposed model determines the optimal number of facilitating vehicles and which mainline vehicles should serve as the facilitating vehicles to cooperatively minimize disruption from ramps. The safety and feasibility of the planned vehicle trajectories are guaranteed at any time. We propose an integrated solution algorithm that incorporates an iterative linear programming method into a novel search process based on a necessary condition for optimality that we identify and prove. The algorithm is highly efficient because it enjoys a significantly reduced search space. The proposed approach, consisting of the MINLP model and the solution algorithm, is evaluated under different traffic demands and mainline-ramp demand ratios and real vehicle arrival patterns from the NGSIM dataset. The performance of the proposed method outperforms benchmark CAV control algorithms, and the computational efficiency is promising for real-time automated merging tasks.

Emergency vehicle preemption (EVP) aims to provide the right-of-way to emergency vehicles (EVs) so that they can travel to the incident location efficiently. The travel time of EVs is the most important indicator of EVP efficiency, which should be minimized by distinct methods or algorithms. However, conventional EVP methods using strobe emitters, light emitters or sirens performs poorly in high-density vehicular traffic. Vehicle-to-everything (V2X) communication plays a pivotal role in intelligent transportation systems (ITS), which can assist EVs to travel safely and efficiently in connected autonomous transport systems (CATS). Enabled by V2X, this paper proposes a deep reinforcement learning-aided mobility-aware lane change algorithm (DRL-MLC) to enhance the efficiency of EVP. In the first stage, the EV learns to change lane based on a policy-based deep reinforcement learning (DRL) algorithm to find the shortest trajectory. In the second stage, autonomous vehicles (AVs) perform mobility-aware lane changing (MLC) to make way for the EV based on the emergency messages (EM) they received. Note that the performance of DRL-MLC strongly relies on the quality of service (QoS) of V2X, and improper network parameters of the on-board units (OBUs) that do not match with the vehicular density will significantly degrade the QoS. Therefore, in the third stage, the proposed algorithm fine-tunes specific parameters including communication range, carrier sensing range, packet rate, and contention window according to the real-time vehicular density based on a curve-fitting optimization method. Our results indicate that at medium-to-high density (e.g., 0.15 veh/m), the average speed of DRL-MLC has more than 49% average improvement than traditional lane changing models, and the ten-minute target travel time for EVs can be achieved by 95% with the proposed algorithm.

The rapid development of electric vehicles raises a higher requirement for charging station operation and management. Therefore, this work proposes a data-driven method aimed at enhancing the economical operation of charging stations. Considering the privacy of charging data and the influence of traffic flow, this data-driven method simulates the spatiotemporal charging demand based on the predicted traffic flow. To obtain precise prediction, this work develops an efficient gated-attention spatiotemporal convolutional network (GSTCN) to explore the long-term spatial and temporal dependence of traffic flow. GSTCN is constructed by two main components: a spatial gated attention (SGA) unit and a temporal gated (T-Gated) attention layer. The spatial pattern of traffic flow is unearthed by the well-designed SGA unit, while the temporal correlation between different time steps is captured through the proposed T-Gated attention layer. Then, an energy storage system (ESS) is employed to improve the effective management of charging stations. Numerical results demonstrate the efficiency of GSTCN in charging station management. GSTCN can produce more accurate prediction data, which leads to a more economical operation of the energy storage system compared with the benchmarks.

We propose a simple and novel method to actively resolve the temporary congestion caused by a freeway mainline vehicle’s facilitating maneuver of creating a gap for on-ramp merging vehicles, under under-critical mainline conditions. We first present an analytical finding derived from the kinematic wave model with a triangular fundamental diagram. That is, when the prevailing mainline traffic is under-critical, the total delay of the mainline vehicles affected by the gap creation does not depend on the choice of speed by which the gap is created, namely the facilitating speed. This is because the recovery wave speed is constantly equal to the characteristic wave speed of the congestion regime,$-w$. In light of this observation, to improve mainline traffic efficiency, we propose the strategy of active congestion resolving, which features a recovery wave of a speed higher than$-w$. Characterizing CAVs’ car-following behaviors by Newell’s simplified car-following model, we analytically show that such a recovery wave can be achieved by properly modifying the values of the affected CAVs’ car-following characteristic parameters and adopting the modified values in a proper way during congestion resolving. Then, in the presence of this active congestion resolving strategy, an optimization program is formulated to seek an optimal facilitating speed that can balance between traffic efficiency and speed variation. Simulation experiments are conducted to validate the effectiveness of the proposed method.

Constructing the prediction intervals (PIs) for electric vehicle (EV) charging demand based on traffic flow information is crucial for the efficient operation of EV charging stations. However, due to the volatile nature of traffic flow, obtaining long-term traffic flow information (e.g., one week in advance) is challenging, particularly for multiple neighbored sites. To address this issue, this work establishes a deep learning prediction framework called Spatiotemporal Periodic Network (STPNet), which utilizes an encoder-decoder architecture. The STPNet incorporates a spatiotemporal and periodic pattern learning technique, and leverages the advantages of convolutional long short-term memory units (ConvLSTM) to quantify the uncertainty of traffic flow. Furthermore, to improve the performance of traffic flow prediction, a spatiotemporal series decomposition strategy based on Seasonal and Trend decomposition using Loess (STL) is employed, and a spatiotemporal PI performance-based loss function is creatively developed in this work. Then, the PIs of the EV charging demand are obtained based on the predicted traffic flow information and an M/M/C/K queuing model. Validated using a real-world dataset, the proposed model has been demonstrated to exhibit effectiveness in generating high-quality EV charging demand PIs for multiple locations.

Accurate long-term origin-destination demand (OD) prediction can help understand traffic flow dynamics, which plays an essential role in urban transportation planning. However, the main challenge originates from the complex and dynamic spatial-temporal correlation of the time-varying traffic information. In response, a graph deep learning model for long-term OD prediction (ST-GDL) is proposed in this paper, which is among the pioneering work that obtains both short-term and long-term OD predictions simultaneously. ST-GDL avoids the conventional multi-step forecasting and thus prevents learning from prediction errors, rendering better long-term forecasts. The proposed method captures time attributes from multiple time scales, namely closeness, periodicity, and trend, to study the features with temporal dynamics. Besides, two gate mechanisms are introduced over the vanilla convolution operation to alleviates the error accumulation issue of typical recurrent forecast in long-term OD prediction. A method based on graph convolution is proposed to capture the dynamic spatial relationship, which projects the transportation network into a graphical time-series. Finally, the long-term OD prediction results are obtained by combining the extracted spatio-temporal features with external features from the meteorological information. Case studies on a practical dataset show that the proposed model is superior to existing methods in long-term OD prediction problems.

Comparative analysis of multimodal travel demand can help transport planners to improve the sustainability of a transport system. This research proposes a framework to qualitatively compare and analyze multimodal travel demand (of constrained and free transport modes) to identify opportunities and prioritize areas for improvement of constrained mode usage. The framework includes two methods. First method, CLAN is a Coarser Level ANalysis, comparing travel patterns and demand ratios of the two modes. Second method, FLAN is a Finer Level ANalysis based on density distributions of zones and OD pairs. The proposed framework is applied to the car (relatively free mode) and transit (constrained mode) OD matrices developed from observed Bluetooth and smart card data, respectively, for the Brisbane City Council region, Australia. The gaps in transit service usage are identified at different sections of the network using both methods. The findings from this study reveal that CLAN can be effectively used to prioritize OD pairs for transit improvement at a coarser level. The OD pairs with the highest and least priority from CLAN are further tested using FLAN. The results from FLAN further confirmed the findings from CLAN. The study showed that both methods have their own advantages and disadvantages if applied independently. However, when applied together, the two methods can help prioritize zones and OD pairs for wider benefits of transport systems such as improvement in transit patronage. The methodological framework proposed in this study is generic and can be applied to compare other multimodal combinations.

Traditional bi-level origin-destination (OD) matrix estimation process adjusts the matrix (at the upper level) based on the deviation between the observed and simulated traffic counts. The problem is mathematically under-determined, and the quality of the solution can be enhanced by restricting the upper level search space with information from other sources. This paper presents a methodology that assimilates sub-path flows in the upper level objective function. The contributions of the study are two-fold: first, it proposes the idea of “structural comparison of sub-path flows” to relax the requirement of “known” penetration rate of vehicles’ trajectories; second, it proposes an innovative upper level formulation where the structural difference between the observed and assigned sub-path flows is integrated with the traditional deviations between the observed and assigned link flows. The sub-path flows can be estimated from advanced data sources such as Bluetooth MAC scanner. The proposed methodology is tested using simulation on a realistic network from Brisbane, Australia and results indicate its practical relevance for situations when the penetration rate of Bluetooth trajectories is generally unknown. The proposed method has a better ability to maintain structural consistency and showed considerable improvements in the quality of OD estimates as compared to the traditional traffic counts-based approach.

This paper proposes a vehicle trajectory planning method for automated on-ramp merging. Trajectory planning tasks of an on-ramp merging vehicle and a mainline facilitating vehicle are formulated as two related optimal control problems. Rather than specifying the merge point via external computational procedures, the location and time that the on-ramp vehicle merges into the mainline are determined endogenously by the optimal control problem of the facilitating vehicle. Bounds on vehicle acceleration are explicitly considered. The Pontryagin Maximum Principle is applied to find the solutions of the optimal control problems. In order to accommodate the constantly changing external environment, the proposed optimal control method is subsequently implemented in a recursive planning framework. Because of the nature of the problem, the length of the planning horizon is time-varying, unlike the conventional model predictive control applications where the planning horizon is a fixed length. Numerical experiments are conducted to study the performances of the proposed methodology under the influence of different leading vehicle trajectories and with different lengths of the planning updating interval. In particular, an experiment involving a real-world leading vehicle trajectory and considering different traffic demand levels are presented. The proposed methodology performs well in these experiments and has demonstrated good potential in real-time applications.

Bluetooth sensors have recently been developed throughout the world for traffic information gathering. Primarily designed for travel time analysis, this article presents a method for vehicular trajectories retrieval. After a short description of some of the challenges at hand in using Bluetooth data in an urban network, a procedure to extract trip information from such data is proposed. It is further analyzed and illustrated at work on a real dataset collected in Brisbane. Last, this article shows that using spatially constrained shortest path analysis, this trip information, once extracted, can be used for the reconstruction of the trajectories. The performance of the process is assessed using both a simulated dataset and one from the real-world acquired in Brisbane, showing encouraging results, with up to 84% of accurately recovered trajectories.

In transport networks, Origin-Destination matrices (ODM) are classically estimated from road traffic counts whereas recent technologies grant also access to sample car trajectories. One example is the deployment in cities of Bluetooth scanners that measure the trajectories of Bluetooth equipped cars. Exploiting such sample trajectory information, the classical ODM estimation problem is here extended into a link-dependent ODM (LODM) one. This much larger size estimation problem is formulated here in a variational form as an inverse problem. We develop a convex optimization resolution algorithm that incorporates network constraints. We study the result of the proposed algorithm on simulated network traffic.

Loop detectors are widely used on the motorway networks where they provide point speed and traffic volumes. Models have been proposed for temporal and spatial generalization of speed for average travel time estimation. Advancement in technology provides complementary data sources such as Bluetooth Media Access Control (MAC) Scanner (BMS), detecting the MAC ID of the Bluetooth devices transported by the traveler. Matching the data from two BMS stations provides individual vehicle travel time. Generally, on the motorways, loops are closely spaced, whereas BMSs are placed a few kilometers apart. In this paper, we fuse BMSs and loops data to define the trajectories of the Bluetooth vehicles. The trajectories are utilized to estimate the travel time statistics between any two points along the motorway. The proposed model is tested using simulation and validated with real data from Pacific Motorway, Brisbane. Comparing the model with the linear-interpolation-based trajectory provides significant improvements.

Transit passenger market segmentation enables transit operators to target different classes of transit users for targeted surveys and various operational and strategic planning improvements. However, the existing market segmentation studies in the literature have been generally done using passenger surveys, which have various limitations. The smart card (SC) data from an automated fare collection system facilitate the understanding of the multiday travel pattern of transit passengers and can be used to segment them into identifiable types of similar behaviors and needs. This paper proposes a comprehensive methodology for passenger segmentation solely using SC data. After reconstructing the travel itineraries from SC transactions, this paper adopts the density-based spatial clustering of application with noise (DBSCAN) algorithm to mine the travel pattern of each SC user. An a priori market segmentation approach then segments transit passengers into four identifiable types. The methodology proposed in this paper assists transit operators to understand their passengers and provides them oriented information and services.