Transportmetrica A-Transport Science最新文献

Accurate short-term passenger demand origin-destination (OD) matrix prediction contributes to the coordination of traffic supply and demand. This study proposes a novel generative adversarial network (GAN) named Conditional Wasserstein Generative Adversarial Network with Gradient Penalty (CWGAN-GP) to predict the network-wide ride-sourcing passenger demand OD matrix. The proposed CWGAN-GP model can not only capture internal spatiotemporal features of OD matrices, but also characterise external dependencies of OD matrices on conditional information, such as the traffic zone-based average traffic speeds, the traffic zone area, and time variables. Based on the ride-sourcing GPS trajectories from Didi Chuxing, Chengdu, China, and ride-sourcing data from the New York City, numerical results illustrate that the predicted OD matrices are in good agreement with the actual ones, and CWGAN-GP has good convergence performance by analysing the discriminator loss and the Wasserstein distance with respect to training epochs. Comparison results also validate the outperformance of CWGAN-GP compared with the other counterpart prediction methods and the reasonability of specific structures of CWGAN-GP. Thus, CWGAN-GP is concluded to be promising to predict network-wide ride-sourcing passenger demand OD matrices.

Inter-business networks are important components of business dynamics and spatial clusters. Extensive literature proclaims the existence of a variety of networks and benefits of diversity within. However, the literature falls short to exhibit the subsistence of connections between economic activities which invariably contribute towards entry decisions of business establishments. In this paper, using National Establishment Time Series (NETS) database, we analyse establishments entering the region around U.S.-400, largely in Kansas, across a period of 20 years (1992–2011). The establishments are categorised into 20 industrial sectors based on their economic activities. Using spatial econometric models and partial least squares regression, yearly relationships (attraction/impedance) among these categories are estimated. The results provide evidence that relationships amidst pre-existing and new-entrants are indeed significant in business entry. This can be critical in understanding and rationale of freight vehicle flows and guide towards corridor utilisation aspects. The inferred relationships could be used to understand how alterations in one sector may perturb other sectors and to inform about upgrading regional attraction, retention and growth-related policies.

Traffic-light modelling is a complex task, because many factors have to be taken into account. In particular, capturing all traffic flows in one model can significantly complicate the model. Therefore, several realistic features are typically omitted from most models. We introduce a mechanism to include pedestrians and focus on situations where they may block vehicles that get a green light simultaneously. More specifically, we consider a generalisation of the Fixed-Cycle Traffic-Light (FCTL) queue. Our framework allows us to model situations where (part of the) vehicles are blocked, e.g. by pedestrians that block turning traffic and where several vehicles might depart simultaneously, e.g. in case of multiple lanes receiving a green light simultaneously. We rely on probability generating function and complex analysis techniques which are also used to study the regular FCTL queue. We study the effect of several parameters on performance measures such as the mean delay and queue-length distribution.

Pedestrian injury in pedestrian-vehicle crash is significantly related to the driver, pedestrian, vehicle, crash and environment characteristics. Driver’s route familiarity has been found greatly associated with driving behaviours. Two-year pedestrian-vehicle crash data in Yunnan Province were studied to investigate the factors that affect pedestrian-injury severities in crashes with familiar and unfamiliar drivers by employing mixed logit models. Eight variables were found significant only in the familiar driver model. And six variables were found significant only in the unfamiliar driver model. Estimation findings indicate that the factors of early morning and sunny weather condition will be better modelled as random parameters in the model for familiar drivers and the same with the factors of rainy weather condition and afternoon peak in the model for unfamiliar drivers. Some more effective and targeted countermeasures are put forward for familiar drivers, unfamiliar drivers and transportation managers to reduce pedestrians’ injury severities.

Delay is one of the most important traffic signal performance measures. In coordinated networks, understanding the characteristics of vehicle arrivals is important for coordination purposes and to properly estimate delays. When observed on a cyclical basis in real-time, distinctive arrival patterns can lead to similar delays, which may go undetected by contemporary delay models. This study proposes a set of enhancements to the Incremental Queue Accumulation (IQA) delay model to overcome the limitations of current models. Additionally, this study proposes a hybrid signal performance measure that combines delay and arrival patterns to depict signal performance truthfully. The enhancements to IQA are realised through an algorithm for the identification of distinctive vehicle arrival groups based on high-resolution signal and detection data. The results demonstrate that the proposed model provides reliable delay estimates (MAPE score in range 4.3–11.2%) while reporting a number of traffic arrival characteristics that are not available from the benchmarked models.

Integrating complicated travel behaviour mechanisms into transportation studies is necessary for understanding and modelling urban mobility. However, insufficient research has been conducted in this direction, especially when travellers make decisions using different mechanisms. This study develops a data-driven framework to model day-to-day route choice dynamics, in which different interpretable travel decision-making mechanisms and efficient model training algorithms are incorporated. The route choice is estimated following a Dirichlet distribution. By introducing a high-order hidden Markov state model, the framework can detect the routine and sudden changes of the mechanism and apply them accordingly for prediction. We propose a particle-based Markov chain Monte Carlo algorithm to estimate model parameters. As a pioneering work that links transportation data with different behaviour mechanisms, we demonstrate the feasibility of the proposed framework through a numerical example. With more transportation data, the proposed approach could become an attractive alternative to conventional transportation models.

Delivery drones are a disruptive technology that is spurring logistics system change, such as the adoption of urban micro-fulfilment centres (MFCs). In this paper, we develop and implement a two-stage continuum approximation (CA) model of this disruptive system in a geographic information system. The model includes common CA techniques at a local level to minimise cost, and then these local solutions are used in a second stage regional location-allocation multiple knapsack problem. We then compare the drone MFC system to a traditional delivery-by-van system and investigate potential cost or emissions savings by adjusting time-window demand, logistical sprawl, electric van alternatives, and MFC emissions. Furthermore, we conduct a sensitivity analysis to show that uncertainty in demand and effective storage density both significantly influence the number of MFCs selected and benchmark our model against commercial solvers. This methodology may also be further developed and applied to other new delivery vehicle modes.