Adam Pel’s research while affiliated with Delft University of Technology and other places

This paper explores temporal patterns in travel production using a full month of production data from traffic analysis zones (TAZ) in the (entire) Netherlands. The mentioned data is a processed aggregated derivative (due to privacy concerns) from GSM traces of a Dutch telecommunication company. This research thus also sheds light on whether such a processed data source is representative of both regular and non-regular patterns in travel production and how such data can be used for planning purposes. To this end, we construct normalized matrix (heatmap) representations of weekly hour-by-hour travel production patterns of over 1200 TAZs, which we cluster using K-means combined with deep convolutional neural networks (inception V3) to extract relevant features. A silhouette score shows that three dominant clusters of temporal patterns can be discerned (K=3). These three clusters have distinctly different within-day and day-to-day production patterns in terms of peak period intensity over different days of the week. Subsequently, a spatial analysis of these clusters reveals that the differences can be related to (easily observable) land-use features such as urbanization levels (i.e., Urban, Rural, and mixed-level). To substantiate this hypothesis and the usefulness of this clustering result, we apply an OVR-SMOTE-XGBoost ensemble classification model on the land-use features of the TAZs (i.e., to identify their cluster). The results of our clustering analysis show that given the land-use features, the overall production patterns are identifiable. Further analysis of the mixed-level areas shows a more complex relationship between temporal heterogeneity and spatial characteristics. Population density seems to impose additional uncertainty on the temporal patterns. All in all, feature selection and spatial and temporal discretization play essential roles in identifying the dominant trip production patterns. These findings are directly useful for data-driven estimation and prediction of demand time series. Furthermore, this study provides further insights into people's mobility, relevant for transportation analysis and policies.

Global System for Mobile Communications (GSM) data provides valuable insights into travel demand patterns by capturing people's consecutive locations. A major challenge, however, is how the polling interval (PI; the time between consecutive location updates) affects the accuracy in reconstructing the spatio-temporal travel patterns. Longer PIs will lead to lower accuracy and may even miss shorter activities or trips when not properly accounted for. In this paper, we analyze the effects of the PI on the ability to reconstruct an origin–destination (OD) matrix. We also propose and validate a new data-driven method that improves accuracy in case of longer PIs. The new method first learns temporal patterns in activities and trips, based on travel diaries, that are then used to infer activity-travel patterns from the (sparse) GSM traces. Both steps are data-driven thus avoiding any a priori (behavioral, temporal) assumptions. To validate the method we use synthetic data generated from a calibrated agent-based transport model. This gives us ground-truth OD patterns and full experimental control. The analysis results show that with our method it is possible to reliably reconstruct OD matrices even from very small data samples (i.e., travel diaries from a small segment of the population) that contain as little as 1% of the population’s movements. This is promising for real-life applications where the amount of empirical data is also limited.

This paper presents an efficient solution method for the matrix estimation problem using a static capacity constrained traffic assignment (SCCTA) model with residual queues. The solution method allows for inclusion of route queuing delays and congestion patterns besides the traditional link flows and prior demand matrix whilst the tractability of the SCCTA model avoids the need for tedious tuning of application specific algorithmic parameters. The proposed solution method solves a series of simplified optimization problems, thereby avoiding costly additional assignment model runs. Link state constraints are used to prevent usage of approximations outside their valid range as well as to include observed congestion patterns. The proposed solution method is designed to be fast, scalable, robust, tractable and reliable because conditions under which a solution to the simplified optimization problem exist are known and because the problem is convex and has a smooth objective function. Four test case applications on the small Sioux Falls model are presented, each consisting of 100 runs with varied input for robustness. The applications demonstrate the added value of inclusion of observed congestion patterns and route queuing delays within the solution method. In addition, application on the large scale BBMB model demonstrates that the proposed solution method is indeed scalable to large scale applications and clearly outperforms the method mostly used in current practice.

This paper presents a straightforward extension from a static into a semi dynamic capacity constrained traffic assignment model. The semi dynamic model is more accurate than its static counterpart whilst, unlike dynamic models, maintains the stability and scalability properties required for application on large scale strategic transport model systems. The paper derives expressions for collective losses and average delays on network-, route-, and link-level from both network operators and traveler's perspective using cumulative in-and outflow curves from the model. The accuracy and stability of the model are compared to its static and dynamic counterparts on theoretical model instances. This comparison shows that the size and temporal distribution of queues and collective losses from the semi dynamic and dynamic models are very similar, but that the spatial distribution is different as the former model ignores spillback. The static model does not resemble the other two models on size, temporal nor spatial distribution of queues and collective losses. The comparison also shows that stability is maintained from the static to the semi dynamic TA model, whereas it is broken for the dynamic TA model, rendering only the static and semi dynamic TA models suitable for strategic applications. The scalability of the model and the effects of the empty network assumption that it relaxes are demonstrated on the large scale strategic transport model of the province of Noord-Brabant (the Netherlands) by comparing outcomes to those of its static counterpart. On this model instance, the empty network assumption in the static model causes omission of up to 76% of collective losses in busy periods and 54% when considering the 24h period. It is therefore very likely that the empty network assumption in static TA models influences (policy) decisions based upon queue size and delay related model outcomes on congested networks. With respect to model scalability, the semi dynamic TA model in its current (prototypical) form requires on average 51% more calculation time in time periods with queues, but it is expected that this could greatly be reduced by merging software components into a single code base. Authors argue that, given the substantial amount of collective loss omitted by the static model, the additional calculation time is a worthwhile penalty to pay. 2 1 INTRODUCTION Strategic traffic assignment (TA) models are used to assess the long-term impact on route choices of transport policies and the design and management of transport systems. As road congestion has become a structural problem in ever more regions around the world, TA model accuracy in congested conditions has become more important. Because strategic TA models are used for long term forecasting, their outcomes should represent stable conditions in which travelers have adapted their route choice behavior to the forecasted scenario. Stability conditions in TA models are mostly operationalized by imposing user equilibrium conditions, where research suggests that a duality gap value (DG, the metric most used to measure the level of disequilibrium) of 1E-04 or lower is needed in strategic context (Boyce et al., 2004; Brederode et al., 2016a, 2019; Caliper, 2010; Han et al., 2015a; Patil et al., 2021). Imposing equilibrium conditions on large scale TA models involves iterative solution algorithms that are computationally expensive. For strategic TA models, there is a clear trade-off between stability and computational requirements on the one hand and accuracy on the other hand (Bliemer et al., 2013; Brederode et al., 2019; Flötteröd and Flügel, 2015). For each type of TA model the trade-off is made differently. In this paper, the framework described in (Bliemer et al., 2017) is used to define and classify the level of accuracy for different types of TA models. By only considering equilibrium models, the three dimensional framework from (Bliemer et al., 2017) simplifies into the two dimensional framework depicted in Figure 1. In this framework, the accuracy of TA models is classified by their spatial and temporal assumptions, where static unrestrained TA models are the least accurate, while dynamic capacity and storage constrained TA models are the most accurate. Below the effects of the different spatial and temporal assumptions on the accuracy of TA models are summarized, for a thorough description of the assumptions themselves the reader is referred to (Bliemer et al., 2017).