<rss version="2.0" xmlns:atom="https://www.w3.org/2005/Atom">
  <channel>
    <title>Transport Research International Documentation (TRID)</title>
    <link>https://trid.trb.org/</link>
    <atom:link href="https://trid.trb.org/Record/RSS?s=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" rel="self" type="application/rss+xml" />
    <description></description>
    <language>en-us</language>
    <copyright>Copyright © 2026. National Academy of Sciences. All rights reserved.</copyright>
    <docs>http://blogs.law.harvard.edu/tech/rss</docs>
    <managingEditor>tris-trb@nas.edu (Bill McLeod)</managingEditor>
    <webMaster>tris-trb@nas.edu (Bill McLeod)</webMaster>
    <image>
      <title>Transport Research International Documentation (TRID)</title>
      <url>https://trid.trb.org/Images/PageHeader-wTitle.jpg</url>
      <link>https://trid.trb.org/</link>
    </image>
    <item>
      <title>Exploring spatiotemporal heterogeneity and nonlinear effects in electric vehicle crash risk prediction: A hybrid modeling approach</title>
      <link>https://trid.trb.org/View/2644105</link>
      <description><![CDATA[Electric vehicle (EV)-related risk and uncertainty pose critical challenges for urban traffic management. Fine-grained crash risk prediction at 1 km × 1 km and hour-of-day resolution remains difficult due to rapidly evolving, strongly spatiotemporally heterogeneous crash patterns. Crash risk research spans risk measurement, prediction modeling, and factor selection, with a move toward interpretable nonlinear hybrid methods, yet temporal dynamics and local heterogeneity remain insufficiently modeled. This study addresses these limitations by first constructing a Spatio-Temporal Adaptive Network Kernel Density Estimation (ST-ANKDE) method that combines network-constrained proximity, cyclic time weighting, severity weighting, and adaptive bandwidths, and then developing a Multiscale Geographically and Temporally Weighted Regression–Extreme Gradient Boosting (MGTWR-XGBoost) method to learn local heterogeneity and nonlinear effects. To capture the influence of preceding periods and adjacent grids, the authors introduce temporal and spatial weighted crash risk variables (T-AccRisk and S-AccRisk). These are analyzed alongside road-network density, built-environment variables, socioeconomic variables, and EV-specific infrastructure variables. An empirical case study on 14,818 EV crashes shows that ST-ANKDE effectively captures crash risk dynamics, with a mean value of 6.57, and reveals pronounced spatiotemporal heterogeneity. The results show that MGTWR-XGBoost, enhanced by S-AccRisk and T-AccRisk to capture spatiotemporal dependence, achieves MAE = 1.54 and RMSE = 2.06 and outperforms standalone machine learning and other hybrid methods; road-network density, built-environment features, population density, and EV infrastructure coefficients exhibit significant spatiotemporal heterogeneity. Moreover, SHapley Additive exPlanations (SHAP) further analyzes nonlinear effects. These findings enable grid-level early warning, priority targeting of high-risk periods/locations, and data-driven deployment of enforcement and infrastructure for EV safety management.]]></description>
      <pubDate>Thu, 15 Jan 2026 14:31:01 GMT</pubDate>
      <guid>https://trid.trb.org/View/2644105</guid>
    </item>
    <item>
      <title>Verbesserte Unfallrekonstruktion durch zusätzliche Anknüpfungstatsachen und KI</title>
      <link>https://trid.trb.org/View/2496624</link>
      <description><![CDATA[Auf Basis von Daten der German In-Depth Accident Study (GIDAS) wurde untersucht, welche Verbesserungen der GIDAS-Unfallrekonstruktion durch Erhebung zusätzlicher Anknüpfungstatsachen zu erwarten sind und ob die Unfallrekonstruktion durch Nutzung künstlicher Intelligenz (KI) optimiert werden kann. Zunächst wurde das Vorgehen in der Rekonstruktion von Verkehrsunfällen im Rahmen des GIDAS-Projektes analysiert. In Abhängigkeit der derzeit verfügbaren Informationsquellen wurden sieben verschiedene Rekonstruktionsvarianten gefunden, für die während der Projektlaufzeit Beispielfälle von erfahrenen GIDAS-Rekonstrukteuren bearbeitet wurden. Diese Fälle bildeten die Basis für eine spätere KI-Potenzialabschätzung. Die Untersuchung derzeit in GIDAS standardmäßig erhobener Informationsquellen und der sich daraus ergebenden Anknüpfungstatsachen ergab, dass aktuell für die Bestimmung von etwa der Hälfte der bisherigen Anknüpfungstatsachen nur eine Informationsquelle vorhanden ist. Unter Einbeziehung neuer Informationsquellen, die zur Verbesserung der Rekonstruktion analysiert wurden, konnte dieser Anteil auf ein Viertel reduziert werden. Anhand der drei beispielhaft gewählten neuen Informationsquellen Event Data Recorder (EDR), Reibwertschätzung (NIRA-Board, NIRA Dynamics AB) und Videos von Verkehrsüberwachungskameras wurde für exemplarische Verkehrsunfälle eine Abschätzung des Mehrwertes der zusätzlichen Erhebung dieser neuen Informationsquellen durchgeführt. Im Ergebnis ermöglichen neue Informationsquellen schon vorhandene GIDAS-Kodierungen zu präzisieren oder sogar vorher unbekannte Parameter zu erfassen. Die derzeit in GIDAS verwendeten Variablen für Toleranzangaben im Rekonstruktionsprozess und ihre Verwendungshäufigkeit waren Teil einer weiteren Analyse. Darauf aufbauend wurde die Genauigkeit der Angabe von rekonstruierten Bewegungsparametern untersucht und für einen Teil der im Projektzeitraum rekonstruierten Unfälle der relative Fehler der Kodierungen für Geschwindigkeitswerte in den verschiedenen Phasen eines Unfalls analysiert. Im Ergebnis zeigt sich, dass Geschwindigkeitsangaben der In-Crash-Phase bei Auslaufbeginn die größten relativen Fehler haben. Die Analysen zum derzeitigen GIDAS-Rekonstruktionsprozess wurden mit einer Befragung von Rekonstrukteuren abgeschlossen, durch die subjektive Beurteilungsfehler im Rekonstruktionsprozess herausgearbeitet wurden. Zwischen den rekonstruierenden Personen ergaben sich teilweise große Unterschiede in der Variablenbestimmung. Ein GIDAS-Teildatensatz von 1.837 Pkw-Pkw-Unfällen ermöglichte grundlegende Betrachtungen für die Prüfung, ob die Verkehrsunfallrekonstruktion durch KI-Methoden effizienter und mit weniger Toleranzen behaftet durchgeführt werden kann. Fünf verschiedene Modelle des maschinellen Lernens wurden trainiert und evaluiert. Bei der Anwendung auf einen GIDAS-Testdatensatz zeigte sich, dass das CatBoost-Modell die höchste Vorhersagegenauigkeit für die Parameter Ausgangsgeschwindigkeit v0, Kollisionsgeschwindigkeit vk, vektorielle Geschwindigkeitsdifferenz delta-v und den Energy-Equivalent-Speed (EES) hatte. Dass KI-Anwendungen einen potenziellen Nutzen für die Unfallrekonstruktion haben, wurde durch Ergebnisse für zwanzig Beteiligte der im Projektzeitraum rekonstruierten Pkw-Pkw-Unfälle deutlich. Bei zwölf beteiligten Pkw (60%) weist die Vorhersage der Ausgangsgeschwindigkeit eine Abweichung von maximal 20% zum vom Rekonstrukteur festgelegten Wert auf; vier Beteiligte (20%) haben maximal nur 10% Abweichung. Die vorhergesagte Ausgangsgeschwindigkeit liegt für fünf Beteiligte (25%) innerhalb des vom Rekonstrukteur angegebenen Toleranzbereichs, der für diese Beteiligten Abweichungen bis zu 20% zulässt. Bei der Arbeit mit den KI-Modellen wurden Schwächen deutlich, deren Ursprung in der relativ kleinen Trainingsdatenmenge vermutet wird. Die entstandenen KI-Modelle sind daher als Ausgangspunkt für den iterativen Prozess der Implementierung und Integration von KI in der Unfallrekonstruktion einzuordnen. ABSTRACT IN ENGLISH: Using data from the German In-Depth Accident Study (GIDAS), it was analyzed which improvements can be expected by interrogating additional sources of information and whether artificial intelligence (AI) allows for optimizing accident reconstruction. First, current accident reconstruction procedures performed as part of the GIDAS project were studied. Seven different reconstruction methods were defined depending on the availability of current sources of information, and multiple example cases were processed by experienced GIDAS reconstructionists for each of these methods during the project duration. These cases also provided the basis for the subsequent estimation of their AI potential. The standard sources of information currently collected during GIDAS data acquisition were documented and their resulting connecting facts examined. Approximately half of those connecting facts are based on only one specific source of information. This proportion was reduced to approximately a quarter by integrating new sources of information. Using three new sources of information as examples Event Data Recorder (EDR), estimation of the coefficient of friction (NIRA-Board, NIRA dynamics AB) and recordings from traffic enforcement cameras), it was possible to evaluate the added value of accessing additional sources of information during traffic accident reconstruction for example cases. As a result, new sources of information allowed to define already existing GIDAS data more precisely or even capture previously unknown parameters. GIDAS variables that define the uncertainty of parameters during the reconstruction process as well as their respective usage frequencies were analyzed. Based on this, the accuracy of reconstruction parameters describing vehicle motion were studied. Data from example cases were used to calculate the relative error of speed values depending on accident phases. As a result, speed values of the in-crash phase at vehicle separation have the highest relative error. The analyses of the current reconstruction methods were completed by interviewing multiple reconstructionists with regard to subjective considerations during variable assessment. The importance assigned to parameters for variable assessment varied widely between different reconstructionists. A dataset of 1,837 vehicle-to-vehicle GIDAS collisions allowed the assessment whether accident reconstruction can be performed more efficiently and with reduced uncertainty using AI methods. For this process five different machine learning algorithms were developed, trained, and evaluated. After the models were applied to a GIDAS test dataset, the CatBoost-model proved to have the highest prediction accuracy for the parameters initial speed, collision speed, change in velocity (delta-v) and energy equivalent speed (EES). Analysis of the example cases using the AI algorithms showed that AI can be beneficial for accident reconstruction. The predicted initial speed by the AI algorithm diverged by only 20% for twelve collision participants and 10% for four collision vehicles, respectively. The predicted initial speed for five collision participants (25%) was within the same margin of uncertainty as determined by a reconstructionist. The application of AI algorithms showed some limitations, most likely due to the small number of cases within the training dataset. The generated AI algorithms should thus be regarded as a starting point for the future iterative process of implementing and integrating AI into traffic accident reconstruction.]]></description>
      <pubDate>Mon, 03 Feb 2025 11:30:37 GMT</pubDate>
      <guid>https://trid.trb.org/View/2496624</guid>
    </item>
    <item>
      <title>Pedaling through the cityscape: Unveiling the association of urban environment and cycling volume through street view imagery analysis</title>
      <link>https://trid.trb.org/View/2450650</link>
      <description><![CDATA[Cycling behavior significantly contributes to urban sustainability and enhances public health. However, revealing the relationship between the built environment and public cycling volume, particularly at the street scale, and achieving urban bicycle-friendly objectives remains a challenge due to a lack of large-scale quantitative methodologies and variability in estimation techniques. This study introduces a novel approach employing street-view imagery and machine learning technologies (specifically training deep learning models on large datasets) to overcome the limitations of traditional methods characterized by low efficiency and narrow geographic coverage. For the implementation of this method, the authors focus on the correlation between urban built environments and cycling volume using Amsterdam, known as a cycling haven, as a case study. The research identifies a dual interaction between street-level and surrounding greenery, manifesting in collaborative and competitive dynamics that jointly shape cycling volume. Moreover, the application of a 4D framework to assess built environments in relation to urban perceptual qualities shows significant correlations with cycling volume. To foster the development of bicycle-friendly cities and enhance public cycling practices, policymakers and urban planners may need to pay greater attention to multidimensional interventions in urban environments.]]></description>
      <pubDate>Mon, 02 Dec 2024 12:49:31 GMT</pubDate>
      <guid>https://trid.trb.org/View/2450650</guid>
    </item>
    <item>
      <title>Full-scale spatio-temporal traffic flow estimation for city-wide networks: a transfer learning based approach</title>
      <link>https://trid.trb.org/View/2289689</link>
      <description><![CDATA[The full-scale spatio-temporal traffic flow estimation/prediction has always been a hot spot in transportation engineering. The low coverage rate of detectors in transport networks brings difficulties to the city-wide traffic flow estimation/prediction. Moreover, it is difficult for traditional analytical traffic flow models to deal with the traffic flow estimation/prediction problem over urban transport networks in a complex environment. Current data-driven methods mainly focus on road segments with detectors. An instance-based transfer learning method is proposed to estimate network-wide traffic flows including road segments without detectors. Case studies based on simulation data and empirical data collected from the open-source PeMS database are conducted to verify its effectiveness. For the traffic flow estimation of segments without detectors, the mean absolute percentage error (MAPE) is approximately 11% for both datasets, which is superior to the existing methods in the literature and reduces MAPE by two percentage points.]]></description>
      <pubDate>Thu, 30 Nov 2023 10:45:16 GMT</pubDate>
      <guid>https://trid.trb.org/View/2289689</guid>
    </item>
    <item>
      <title>Traffic conflict prediction using connected vehicle data</title>
      <link>https://trid.trb.org/View/2138632</link>
      <description><![CDATA[Transportation safety studies have been mostly focused on using crash data that are rare events. Alternatively, conflict estimation can be used to assess safety. This has been proven as a proactive design methodology that does not rely on crashes and requires shorter observation. Traditionally, the safety studies involving both these reactive and proactive methods were based on aggregated data that does not take individual vehicle dynamics into consideration. This paper addresses this research gap by proposing a novel real-time conflict prediction methodology that uses previous instance trajectory data of individual vehicles to understand whether there can be potential conflict in the near future. A long-short term memory (LSTM) model is developed that can apprehend a conflict situation 9 s in the future. Data from connected vehicles have been used. The proposed model returned a recall of 81% with a false alarm rate of 28%. The predictive model has the potential to be implemented on vehicle dashboards to warn drivers of a conflict. The authors have also used SHAP (SHapley Additive exPlanation) to interpret the results from the LSTM model. It was deduced that acceleration above 0.3 m/s², deceleration within −1.5 m/s² to −0.25 m/s², and speed of more than 40kph were responsible for inducing a conflict.]]></description>
      <pubDate>Mon, 24 Apr 2023 09:51:03 GMT</pubDate>
      <guid>https://trid.trb.org/View/2138632</guid>
    </item>
    <item>
      <title>Towards safe autonomous driving : from predictive safety evaluation to monitoring of neural networks</title>
      <link>https://trid.trb.org/View/2075191</link>
      <description><![CDATA[Autonomous driving is expected to bring several benefits, in particular regarding safety. This thesis aim to contribute towards two questions concerning safety: "What is the potential safety benefit of autonomous driving?'' and "How can we ensure safe operation of such vehicles?''. In the first part of the thesis, methods for evaluating the safety benefit are investigated. In particular predictive effectiveness evaluation based on resimulation of accident data, using models to estimate new outcomes in case the safety system had been available. To illustrate the methodology, four examples of gradual increase in model complexity are presented. First, an Autonomous Emergency Braking (AEB) system using a sensor model, decision algorithm, vehicle dynamics model and regression based injury model. This is extended in a Forward Collision Warning (FCW) system which additionally requires a driver model to simulate driver reactions. The third example shows how an active, AEB, and passive, airbag, system can be combined. Finally the fourth example combines several systems to emulate a highly automated vehicle. Apart from predicting the real world performance, this analysis also identifies current safety gaps by studying the residual of the accident set. Safety benefit estimation using accident data gives an evaluation on the current accident distributions, however, the systems may introduce new accidents if not operated as intended. In the second part of the thesis, safety verification processes with the intent of preventing unsafe operation, are presented. This is particularly challenging for machine learning based components, such as neural networks. In this case, traditional analytical verification approaches are difficult to apply due to the non-linearity and high dimensional parameter spaces. Similarly, statistical safety arguments often require unfeasible amounts of annotated validation data. Instead, monitor functions are investigated as a complement to increase safety during operation. The method presented estimates the similarity of the driving environment, compared to the training data, where decisions inferred from novel data can be considered less reliable. Although not providing a complete safety assurance, the methodology show promising initial results for increasing safety. In addition, it could potentially be used to collect novel data and reduce redundancy in training data.]]></description>
      <pubDate>Fri, 02 Dec 2022 11:40:44 GMT</pubDate>
      <guid>https://trid.trb.org/View/2075191</guid>
    </item>
    <item>
      <title>Traffic dynamics estimation by using raw floating car data</title>
      <link>https://trid.trb.org/View/1725936</link>
      <description><![CDATA[Massive datasets of Floating Car Data (FCD) are collected and thereafter processed to estimate and predict traffic conditions. In the framework of short-term traffic forecasting, machine learning techniques have become very popular. However, the big datasets available today contain for the most part easily predictable data, that are data observed during recurrent conditions. Integration of different machine learning techniques with traffic engineering notions must contribute to obtain new transportation-oriented data-driven methods. In this paper the authors address traffic dynamics estimation by using individual FCD in order to develop an integrative framework able to recognize and select the suitable method for traffic forecasting. Taking into account the spatial distributions of individual FCD positions the authors retrieve a new spatial-based criterion for the integration of models.]]></description>
      <pubDate>Tue, 10 Nov 2020 09:19:59 GMT</pubDate>
      <guid>https://trid.trb.org/View/1725936</guid>
    </item>
    <item>
      <title>Dynamic Tolling of HOT Lanes Through Simulation of Expected Traffic Conditions</title>
      <link>https://trid.trb.org/View/1343454</link>
      <description><![CDATA[Dynamic tolling of High-Occupancy/Toll lanes (HOTL) is a challenging task, particularly given the range of policy constraints that could potentially be applicable. Methodological literature on the topic is relatively sparse or is considered proprietary information, particularly with respect to current implementations. A framework was developed for the Ministry of Transportation of Ontario (MTO) that utilizes micro-simulation of short-term future conditions to evaluate alternative tolling rate strategies and select the rate to be applied for the next time interval. In this case, the objective of the selection algorithm was the choice of a toll rate that maximized utilization of the HOTL subject to maintenance of a specified minimum average speed. Short-term future conditions can be anticipated by supplying the micro-simulation model with traffic flow data collected “upstream” of the HOTL. The framework, as conceived, has a high degree of flexibility. It could be applied on-line or simulation trials conducted off-line could be used to generate a look-up table relating toll rates to traffic conditions. Measurements of actual HOTL performance under the chosen toll rate could be recorded and generalized using a neural network. Reinforcement learning or other machine learning methods could be applied either on or off-line to improve the decision-making process. The process, including the dynamic pricing algorithm but excluding generalization or learning capabilities, was implemented in AIMSUN for evaluation and trial application. The current paper describes the operation of the framework and its application to the evaluation of a hypothetical HOVL to HOTL conversion.]]></description>
      <pubDate>Fri, 13 Feb 2015 16:22:08 GMT</pubDate>
      <guid>https://trid.trb.org/View/1343454</guid>
    </item>
    <item>
      <title>Estimation of Traffic Dynamics Models with Machine-Learning Methods</title>
      <link>https://trid.trb.org/View/776994</link>
      <description><![CDATA[Speed–density relationships are a classic way of modeling stationary traffic relationships. Besides offering valuable insight into traffic stream flows, such relationships are widely used in dynamic traffic assignment (DTA) systems. In this research, an alternative paradigm for traffic dynamics models, appropriate for traffic simulation models and based on machine-learning approaches such as k-means clustering, k-nearest-neighborhood classification, and locally weighted regression is proposed. Although these models may not provide as much insight into traffic flow theory as speed–density relationships do, they allow for easy incorporation of additional information to speed estimation and hence may be more appropriate for use in DTA models, especially simulation-based models. This paper (with data from a network in Irvine, California) demonstrates that such machine-learning methods can considerably improve the accuracy of speed estimation.]]></description>
      <pubDate>Thu, 01 Jun 2006 08:09:11 GMT</pubDate>
      <guid>https://trid.trb.org/View/776994</guid>
    </item>
  </channel>
</rss>