Transportation Research Part C-Emerging Technologies最新文献

Data provided by floating vehicles have the same limitation as that provided by an imperfect stationary detector that undercounts vehicles: they provide data from an unknown proportion of the total flow. Since the knowledge of this proportion is critical for some applications, its value is usually estimated by complementing with data provided by other monitoring infrastructure. However, this type of infrastructure is not the norm in many cities, particularly in developing countries. This paper extends a previously proposed method to estimate this penetration solely from a series of inter-detection times. The improved method is based on Bayesian inference and supports a series of passage times coming from multilane infrastructure, albeit still requiring an uninterrupted traffic regime. The method was applied to actual RFID interdetection times from a four-lane weaving urban freeway segment in Santiago, Chile. The resulting estimated rates are consistent with those reported by the freeway operator, showing the method’s capabilities. Thus, the method is instrumental in successfully taking full advantage of technology already in place in many cities without needing new investments in monitoring infrastructure.

Connected and Automated Vehicles (CAVs) are expected to reshape the transportation system, and cooperative group intelligence of CAVs has great potential for improving transportation efficiency and safety. One challenge for CAV group driving is the decision-making under scenarios mixed with CAV and human-driven vehicles (HDV). Current studies mainly use methods based on single physical rules such as platoon driving or formation switch control, failing to reach a balanced and homogeneous state of optimal efficiency and risk in mixed traffic environments. In addition, most studies focus only on one specific type of scene, lacking the scene adaptability to various surrounding conditions. This paper proposes a homogeneous multi-vehicle cooperative group decision-making method targeting mixed traffic scenarios. A bi-level framework composed of behavior-level and trajectory-level decision-making is established to achieve balanced optimal cooperation. A region-driven behavioral decision mechanism is designed to decompose vehicle actions into a unified form of sequential target regions. Solutions are derived based on Cooperative Driving Safety Field, a risk assessment module inspired by field energy theory. The trajectory-level decision module takes the target regions as input and generates the control quantities of the CAVs through target point selection, conflict reconciliation, and dynamic constraint consideration. Experimental results on 19 various scenarios and continuous traffic flow scenes indicate that the proposed method significantly increases passing efficiency, reduces driving risk, and improves scene adaptability. In addition, experiments on multiple kinds of scenarios including intersections, ramps, bottlenecks, etc. prove that our method can adapt to various road topology structures. Feasibility is also verified through scaled physical platform validations and real-vehicle road tests.

Carsharing operators (CSOs) are adapting their service over time to meet changing demands and grow their market share. Service areas are enlarged and, in some cities, “dual-mode settings” evolve, incorporating free-floating carsharing (FFcs) as a new service alongside existing station-based carsharing (SBcs). This paper proposes a methodology to discuss adoption dynamics in such a context, specifically focusing on the impact of existing services and service extensions on the adoption of the new service. We propose a framework, comprising of two parts: a potential market assessment and an adoption model. The potential market assessment focuses on establishing the relationships between the local population, carsharing memberships and Points of Interest (POIs) within the given service area. The adoption model then describes the likelihood of consumers adopting the FFcs service. By combining these two models, the effects of service extensions can be assessed. We evaluate the framework using a nearly six year dataset from Communauto, Montreal. The first 35 months of data are set as training data, while the subsequent 33 months are used for validation of predictive performance. Results demonstrate that the proposed model accurately predicts adoption dynamics. Prior experience of SBcs and initial information spread are found to be key parameters for demand prediction determining early adoption peaks and, due to follower effects, also impact long-term demand. Additionally, we quantify the importance of covering residential areas and points of interests in the service area, highlighting the synergy effects of service area expansions.

Confronted with growing and fluctuating traffic demands, urban transportation system has been encountering mounting challenges in traffic congestion, especially at intersections. With enhanced traffic control precision enabled by the emerging Connected and Automated Vehicle (CAV) technologies, this study proposes a hybrid control strategy for connected and automated traffic at urban intersections, which enables the integration of diverse control schemes to harness their strengths and mitigate their weaknesses. With rolling horizon strategy, a nonlinear optimization model is developed to determine the optimal traffic control plans considering both current status and forthcoming vehicle arrivals. Vehicle delays are elaborately characterized without relying on any empirical assumptions. The original model is converted to a Mixed Integer Programming with Quadratic Constraints (MIP-QC) by employing appropriate linearization techniques, which could be solved by commercial solvers. For the acquisition of instant and reliable solutions, a multilayer feedforward network-based approximate algorithm is developed, referred as Value Approximation Control (VAC) algorithm. Theoretical derivation is provided to validate the capability of VAC algorithm in the precise approximations of the value function in the traffic plan optimization problem, and ultimately enabling to acquire global optimal solutions via specific network design and training techniques. Numerical experiments on both artificial and researcher-collected datasets demonstrate that our proposed VAC algorithm achieves performance nearly equivalent to the mathematical model. Significantly, it outperforms current state-of-art traffic control methods in terms of both intersection throughput and average vehicle delay. Moreover, sensitivity analysis reveals the robustness of the VAC algorithm against inaccuracy in vehicle arrival information, and the stable performance even in the presence of significant disturbances.

Understanding travellers’ route preferences allows for the calculation of traffic flow on network segments and helps in assessing facility requirements, costs, and the impact of network modifications. Most research employs logit-based choice methods to model the route choices of individuals, but machine learning models are gaining increasing interest. However, all of these methods typically rely on a single ‘best’ model for predictions, which may be sensitive to measurement errors in the training data. Moreover, predictions from discarded models might still provide insights into route choices. The ensemble approach combines outcomes from multiple models using various pattern recognition methods, assumptions, and/or data sets to deliver improved predictions. When configured correctly, ensemble models offer greater prediction accuracy and account for uncertainties. To examine the advantages of ensemble techniques, a data set from the I-35 W Bridge Collapse study in 2008, and another from the 2011 Travel Behavior Inventory (TBI), both in Minneapolis–St. Paul (The Twin Cities) are used to train a set of route choice models and combine them with ensemble techniques. The analysis considered travellers’ socio-demographics and trip attributes. The trained models are applied to two datasets, the Longitudinal Employer-Household Dynamics (LEHD) commute trips and TBI morning peak trips, for validation. Predictions are also compared with the loop detector records on freeway links. Traditional Multinomial Logit and Path-Size Logit models, along with machine learning methods such as Decision Tree, Random Forest, Extra Tree, AdaBoost, Support Vector Machine, and Neural Network, serve as the foundation for this study. Ensemble rules are tested in both case studies, including hard voting, soft voting, ranked choice voting, and stacking. Based on the results, heterogeneous ensembles using soft voting outperform the base models and other ensemble rules on testing sets.

In a busy railway marshalling station, train and engine traffic management in the receiving/departure yard plays a crucial role in efficient and stable operations. Traditionally, a track assignment problem (TAP) is solved to assign tracks to trains for berthing at a receiving/departure yard. However, the TAP does not encompass shunting operations in the yard (e.g., engine replacement and train disassembly), which can result in additional scheduling challenges for dispatchers and route conflicts between operations. This paper investigates a train and engine routing and scheduling problem (TERSP) in a receiving/departure yard of railway marshalling stations, which involves simultaneously assigning routes and scheduling route-setting start times for both train and shunting operations to be conducted in the yard. By introducing the concepts of task, activity, and pattern, we transform the original problem into assigning pre-generated patterns incorporating both route and route-setting start time alternatives to activities. The transformed problem is formulated into a compact binary integer linear programming model with a linear number of constraints and the objective of minimizing the total time deviation of all involved tasks. An improved technique that relies on listing all maximal (bi)cliques in a constructed graph is designed to effectively model the time coherence and track section occupation constraints. A heuristic that gradually expands the patterns for the identified key activities by adding more start time alternatives is applied to remedy an infeasible model caused by potential route conflicts. In addition, a rolling horizon algorithm that decomposes the original problem into consecutive smaller stages using either a time-rolling or a train-rolling rule is developed to efficiently solve instances. Finally, numerical experiments based on the physical layouts and real timetables of a receiving yard and a departure yard of a large marshalling station in China are conducted to assess the performance and applicability of our proposed approaches. The results demonstrate that our approaches typically find (near-)optimal solutions within several minutes for the investigated instances by simultaneously addressing different classes of yard operations and resources.

The electric vehicle, as a green and sustainable technology, has gained tremendous development and application recently in the logistics distribution system. However, the increasing workload and limited infrastructure capacity pose challenges for electric vehicles in the pickup and delivery operating system, including task allocation, electric vehicle routing, and queue scheduling. To address these issues, this paper introduces a pickup and delivery problem with electric vehicles and time windows considering queues, which considers queue scheduling for multiple electric vehicles when operating at the same site. A novel mixed integer linear programming model is proposed to minimize the cost of travel distance and queue time. An adaptive hybrid neighborhood search algorithm is developed to solve the moderately large-scale problem. Experimental results demonstrate the effectiveness of the model and adaptive hybrid neighborhood search algorithm. The competitive performance of the developed algorithm is further confirmed by finding 9 new best solutions for the pickup and delivery problem with electric vehicles and time windows benchmark instances. Moreover, the results and sensitivity analysis of objective weight costs highlight the impact and importance of considering queues in the studied problem and obtain some management insights.

This paper introduces a novel hybrid car-following strategy for connected automated vehicles (CAVs) to mitigate traffic oscillations while simultaneously improving CAV car-following (CF) distance-maintaining efficiencies. To achieve this, our proposed control framework integrates two controllers: a linear feedback controller and a deep reinforcement learning controller. Firstly, a cutting-edge linear feedback controller is developed by non-linear programming to maximally dampen traffic oscillations in the frequency domain while ensuring both local and string stability. Based on that, deep reinforcement learning (DRL) is employed to complement the linear feedback controller further to handle the unknown traffic disturbance quasi-optimally in the time domain. This unique approach enhances the control stability of the traditional DRL approach and provides an innovative perspective on CF control. Simulation experiments were conducted to validate the efficacy of our control strategy. The results demonstrate superior performance in terms of training convergence, driving comfort, and dampening oscillations compared to existing DRL-based controllers.

Cooperation at unsignalized intersections in mixed traffic environments, where Connected and Autonomous Vehicles (CAVs) and Manually Driving Vehicles (MVs) coexist, holds promise for improving safety, efficiency, and energy savings. However, the mixed traffic at unsignalized intersections present huge challenges like MVs’ uncertainties, the chain reaction and diverse interactions. Following the thought of the situation-aware cooperation, this paper proposes a Reasoning Graph-based Reinforcement Learning (RGRL) method, which integrates a Graph Neural Network (GNN) based policy and an environment providing mixed traffic with uncertain behaviors. Firstly, it graphicly represents the observed scenario as a situation using the interaction graph with connected but uncertain (bi-directional) edges. The situation reasoning process is formulated as a Reasoning Graph-based Markov Decision Process which infers the vehicle sequence stage by stage so as to sequentially depict the entire situation. Then, a GNN-based policy is constructed, which uses Graph Convolution Networks (GCN) to capture the interrelated chain reactions and Graph Attentions Networks (GAT) to measure the attention of diverse interactions. Furthermore, an environment block is developed for training the policy, which provides trajectory generators for both CAVs and MVs. A reward function that considers social compliance, collision avoidance, efficiency and energy savings is also provided in this block. Finally, three Reinforcement Learning methods, D3QN, PPO and SAC, are implemented for comparative tests to explore the applicability and strength of the framework. The test results demonstrate that the D3QN outperformed the other two methods with a larger converged reward while maintaining a similar converged speed. Compared to multi-agent RL (MARL), the RGRL approach showed superior performance statistically, reduced the number of severe conflicts by 77.78–94.12 %. The RGRL reduced average and maximum travel times by 13.62–16.02 %, and fuel-consumption by 3.38–6.98 % in medium or high Market Penetration Rates (MPRs). Hardware-in-the-loop (HIL) and Vehicle-in-the-loop (VehIL) experiments were conducted to validate the model effectiveness.

Transit users’ lifecycle behavior pattern transition reflects the continuous and multi-phase changes in how frequently and regularly users utilize public transit over their lifetime. Predicting transit users’ lifecycle behavior pattern transition is vital for enhancing the efficiency and responsiveness of transportation systems. Thus, this study incorporates lifecycle analysis in predicting long-term sequential behavioral pattern transition processes to go beyond just examining user churning at a single point in time. Specifically, this study proposes the TripChain2RecDeepSurv, a novel model that pioneers the individual-level analysis of lifecycle behavior status transitions (LBST) within public transit systems. The TripChain2RecDeepSurv is composed of (1) the TripChain2Vec module for encoding transit users’ trip chains; (2) the self-attention Transformer module for exploring the latent features related to spatiotemporal patterns; (3) the recurrent deep survival analysis module for predicting LBSTs. We demonstrate TripChain2RecDeepSurv’s predictive performance for empirical analysis by employing Shenzhen Bus data. Our model achieves a 74.39% accuracy rate in churn determination and over 80% accuracy in status sequence identification on the churn path. In addition, our findings highlight the segmented nature of Kaplan-Meier curves and identify the optimal intervention time against the user churning process. Meanwhile, the proposed model provides individual-level heterogeneity analysis, which emphasizes the significance of customizing user engagement strategies, advocating for interventions that extend users’ engagement in patterns with high-frequency transit usage to curb the transition to less frequent travel usage.