Accident de la route | EPFL Graph Search

vignette|Accident de la route à une intersection. thumb|Nettoyage de la chaussée après un accident de voitures frontal sur l'avenue Charles-de-Gaulle à Ouagadougou. Un accident de la route (ou accident sur la voie publique : AVP, ou en France accident de la circulation) est une collision non intentionnelle, qui a lieu sur le réseau routier entre un engin roulant (notamment automobile, camion, moto, vélo) et tout autre véhicule ou personne ou animal, se déplaçant ou non, qui engendre au moins des dégâts matériels, voire des traumatismes ou le décès d'une ou plusieurs personnes impliquées. De nombreux facteurs contribuent aux risques de collision tels que l'état de la chaussée, le volume de circulation, une vitesse de circulation inadaptée, l’environnement routier, la météorologie, l'expérience du conducteur mais aussi son comportement et son état physique. Selon l'OMS, les accidents de la route dans le monde tuent chaque année environ de personnes et font de 25 à de blessés. Il s

Motorway Traffic Risks Identification Model - MyTRIM Methodology and Application

Minh Hai Pham

Road traffic crashes are becoming increasing concerns in many countries. In Europe, many efforts have been devoted to improve road traffic safety yet the important target of halving the number of yearly road deaths in 2010 could not be achieved in many European countries. Among different road types, motorways are safe by design yet crashes if occur would be severe due to high speed practiced. If motorway traffic crash risk could be identified, lives could be saved and severity could be reduced. For this objective, the current thesis aims to establish a methodology for developing models capable of identifying real-time traffic crash risk on motorways. A real-time MotorwaY Traffic Risk Identification Model (MyTRIM) is developed for a study site on motorway A1 in Switzerland. MyTRIM is tested, validated with real data. Three types of historical data altogether available at the study site are used for developing MyTRIM. The data include individual vehicle traffic data collected from double loop traffic detectors, meteorological data from meteorological station located near the study site, and a crash database containing crashes recorded by the police. Based on crash time, pre-crash data representing traffic and meteorological conditions leading to crashes are extracted. Similarly, non-crash data representing traffic and meteorological conditions that are unrelated with crashes are also extracted. As crashes are rare events, a methodology for sampling non-crash data comparable with pre-crash data is developed using clustering – classification basis: non-crash data are clustered into groups; pre-crash data are classified into obtained groups; pre-crash and non-crash data within one group are similar and therefore, comparable. Each group is called a traffic regime. Under each traffic regime, a regime-based Risk Identification Model (RIM) is developed to differentiate pre-crash and non-crash data. Given a new datum, regime-based RIM must be able to classify the datum into pre-crash or non-crash. As a result of the model development, variables which are important for the differentiation are also identified. These important variables can be potential for implementing countermeasures to prevent the risk from ending up with a crash. MyTRIM is developed based on the outputs from regime-based RIM. MyTRIM memorizes the latest risk evolution to predict whether there is crash risk in the coming time interval. Regime-based RIM and MyTRIM are tested and validated using real data. Results show that regime-based RIM and MyTRIM perform with high accuracy. The output of MyTRIM can be useful for traffic operators as an input for actively managing the traffic. The developed methodology can be applied for any motorway traffic sections where similar data are available.

EPFL2011

Risk indicators - Unsafe traffic conditions

Edward Chung, Minh Hai Pham

This study proposes a framework for developing a pro-active system aiming at reducing motorway crash risk based on risk indicators. Although motorways are of safe road type by design, a crash if happens would be really severe. Reducing motorway crash risk would prevent crashes or diminish their severity. To that purpose, risk indicators are tested and validated with real data including traffic data, meteorological data, and crash databases from a study site in Switzerland. A guideline about how to define the risk criteria is proposed and applied in a simulation model which is the platform for implementing a pro-active crash risk prevention system. The model is calibrated to best represent a road section at the study site. An implementation example of preventive measures is introduced and its outcome is evaluated.

COST Office2009

Multivariate time series forecasting for freeway networks

Semin Kwak

This dissertation introduces traffic forecasting methods for different network configurations and data availability.Chapter 2 focuses on single freeway cases.Although its topology is simple, the non-linearity of traffic features makes this prediction still a challenging task.We propose the dynamic linear model (DLM) to approximate the non-linear traffic features. Unlike a static linear regression model, the DLM assumes that its parameters change over time.We design the DLM with time-dependent model parameters to describe the spatiotemporal characteristics of time-series traffic data. Based on our DLM and its model parameters analytically trained using historical data, we suggest the optimal linear predictor in the minimum mean square error (MMSE) sense.We compare our prediction accuracy by estimating expected travel time based on the traffic prediction for freeways in California (I210-E and I5-S) under highly congested traffic conditions with other baselines. We show significant improvements in accuracy, especially for short-term prediction.Chapter 3 aims to generalize the DLM to extensive freeway networks with more complex topologies.Most resources would be consumed to estimate unnecessary spatiotemporal correlations if the DLM was directly used for a large-scale network.Defining features on graphs relaxes such issues by cutting unnecessary connections in advance based on predefined topology information.Exploiting the graph signal processing, we represent traffic dynamics over freeway networks using multiple graph heat diffusion kernels and integrate the kernels into DLM with Bayes' rule. We optimize the model parameters using Bayesian inference to minimize the prediction errors.The proposed model demonstrates prediction accuracy comparable to state-of-the-art deep neural networks with lower computational effort. It notably achieves excellent performance for long-term prediction through the inheritance of periodicity modeling in DLM.Chapter 4 proposes a deep neural network model to predict traffic features on large-scale freeway networks.These days, deep learning methods have heavily tackled traffic forecasting problems of freeway networks because they are outstanding at learning highly complex correlations between variables both in time and space, which the linear models might be limited to.Adopting a graph convolutional network (GCN) becomes a standard to extract spatial correlations; therefore, most works have achieved great prediction accuracy by implanting it into their architecture.However, the conventional GCN has the drawback that receptive field size should be small, i.e., barely refers to traffic features of remote sensors, resulting in inaccurate long-term prediction.We suggest a forecasting model called two-level resolution deep neural network (TwoResNet) that overcomes the limitation.It consists of two resolution blocks:The low-resolution block predicts traffic on a macroscopic scale, such as regional traffic changes.On the other hand, the high-resolution block predicts traffic on a microscopic scale by using GCN to extract spatial correlations, referring to the regional changes produced by the low-resolution block.This process allows the GCN to refer to the traffic features from remote sensors.As a result, TwoResNet achieves competitive prediction accuracy compared to state-of-the-art methods, especially showing excellent performance for long-term predictions.

EPFL2022