Highly Automated Vehicles (AVs) have the capability to revolutionize the transportation system. These systems have the possibility to make roads safer as AVs do not have limitations that human drivers do, many of which are common causes of vehicle crashes (e.g., distraction or fatigue) often defined generically as human error. The deployment of AVs is likely to be very gradual however, and there will exist situations in which the AV will be driving in close proximity with human drivers across the foreseeable future. Given the persistent crash problem in which the makority of crashes are attributed to driver error, humans will continue to create potential collision scenarios that an AV will be expected to try and avoid or mitigate if developed appropriately. The absence of unreasonable risk in an AVs ability to comprehend and react in these situations is referred to as operational safety. Unlike advanced driver assistance systems (ADAS), highly automated vehicles are required to perform the entirety of the dynamic driving task (DDT) and have a greater responsibility to achieve a high level of operational safety. To address this concern, scenario-based testing has increasingly become a popular option for evaluating AV performance. On a functional level, an AV typically consists of three basic systems: the perception system, the decision and path planning system, and vehicle motion control system. A minimum level of performance is needed in each of these functional blocks to achieve an adequate level of operational safety. The goal of this study was to investigate the effects that perception system performance (i.e., target object state errors) has on vehicle operational safety in collision scenarios similar to that created by human drivers. In the first part of this study, recent annual crash data was used to define a relevant crash population of possible scenarios involving intersections that an AV operating as an urban taxi may encounter. Common crash maneuvers and characteristics were combined to create a set of testing scenarios that represent a high iii percentage of the overall crash population. In the second part of this study, each test scenario was executed using an AV test platform during closed road testing to determine possible real-world perception system performance. This provided a measure of the error in object detection measurements compared to the ideal (i.e., where a vehicle was detected to be compared to where it actually was). In the third part of this study, a set of vehicle simulations were performed to assess the effect of perception system performance on crash outcomes. This analysis simulated hypothetical crashes between an AV and one other collision partner. First an initial worst-case impact configuration was defined and was based on injury outcomes seen in crash data. The AV was then simulated to perform a variety of evasive maneuvers based on an adaptation of a non-impaired driver model. The impact location and orientation of the collision partner was simulated as two states: one based on the object detection of an ideal perception system and the other based on the object detection of the perception system from the AV platform used during the road testing. For simulations in which the two vehicles contacted each other, a planar momentum-impulse model was used for impact modeling and injury outcomes were predicted using an omni-directional injury model taken from recent literature. Results from this study indicate that errors in perception system measurements can change the perceived occupant injury risk within a crash. Sensitivity was found to be dependent on the specific crash type as well as what evasive maneuver is taken. Sensitivities occurred mainly due to changes in the principal direction of force for the crash and the interaction within the injury risk prediction curves. In order to achieve full operational safety, it will likely be important to understand the influence that each functional system (perception, decision, and control) may have on AV performance in these crash scenarios. / Master of Science / Highly Automated Vehicles (AVs) have the capability to revolutionize the transportation system. These systems have the possibility to make roads safer as AVs do not have many of the limitations that human drivers do, many of which are common causes of vehicle crashes (e.g., distraction or fatigue). AVs will be expected to drive alongside human drivers, and so these drivers are likely to continue to be at fault in causing crashes. As part of ensuring safety, AVs will reasonably be expected to try and avoid or help reduce the severity of these crashes. AVs operate using three main systems: the perception system which consists of sensors that see the objects around the AV, the decision and path planning system, which makes decision on what the AV will do, and the vehicle motion control system. Due to the nature of the real-world, these systems may not work exactly as intended which may affect the ability of the AV to react to possible crash scenarios. Because of this, the goal of this study was to investigate the effects that perception system performance (i.e., target object state errors) has on the ability of an AV to react to crash scenarios similar to those created by human drivers. This study first defined crash scenarios using real-world crash data. A real-world perception system was then tested in these scenarios to determine object detection performance. Based on this performance, effects on safety were assessed through vehicle crash simulations. Results from this analysis showed that safety can vary based on both perception system performance and crash scenario. This highlights that it will be important to address system performance in order to achieve high levels of driving safety.

Les travaux présentés dans ce mémoire s’inscrivent dans le cadre duprojet PANsafer (Vers un Passage A Niveau plus sûr), lauréat de l’appel ANR-VTT2008. Ce projet est labellisé par les deux pôles de compétitivité i-Trans et Véhiculedu Futur. Le travail de la thèse est mené conjointement par le laboratoire IRTESSETde l’UTBM et le laboratoire LEOST de l’IFSTTAR.L’objectif de cette thèse est de développer un système de perception permettantl’interprétation de scénarios dans l’environnement d’un passage à niveau. Il s’agitd’évaluer des situations potentiellement dangereuses par l’analyse spatio-temporelledes objets présents autour du passage à niveau.Pour atteindre cet objectif, le travail est décomposé en trois étapes principales. Lapremière étape est consacrée à la mise en place d’une architecture spatiale des capteursvidéo permettant de couvrir de manière optimale l’environnement du passageà niveau. Cette étape est mise en oeuvre dans le cadre du développement d’unsimulateur d’aide à la sécurité aux passages à niveau en utilisant un système deperception multi-vues. Dans ce cadre, nous avons proposé une méthode d’optimisationpermettant de déterminer automatiquement la position et l’orientation descaméras par rapport à l’environnement à percevoir.La deuxième étape consisteà développer une méthode robuste de suivi d’objets enmouvement à partir d’une séquence d’images. Dans un premier temps, nous avonsproposé une technique permettant la détection et la séparation des objets. Le processusde suivi est ensuite mis en oeuvre par le calcul et la rectification du flotoptique grâce respectivement à un modèle gaussien et un modèle de filtre de Kalman.La dernière étape est destinée à l’analyse des trajectoires 2D reconstruites parl’étape précédente pour l’interprétation de scénarios. Cette analyse commence parune modélisation markovienne des trajectoires 2D. Un système de décision à basede théorie de l’évidence est ensuite proposé pour l’évaluation de scénarios, aprèsavoir modélisé les sources de danger.L’approche proposée a été testée et évaluée avec des données issues de campagnesexpérimentales effectuées sur site réel d’un passage à niveau mis à disposition parRFF. / The main objective of this thesis is to develop a system for monitoringthe close environment of a level crossing. It aims to develop a perception systemallowing the detection and the evaluation of dangerous situations around a levelcrossing.To achieve this goal, the overall problem of this work has been broken down intothree main stages. In the first stage, we propose a method for optimizing automaticallythe location of video sensors in order to cover optimally a level crossingenvironment. This stage addresses the problem of cameras positioning and orientationin order to view optimally monitored scenes.The second stage aims to implement a method for objects tracking within a surveillancezone. It consists first on developing robust algorithms for detecting and separatingmoving objects around level crossing. The second part of this stage consistsin performing object tracking using a Gaussian propagation optical flow based modeland Kalman filtering.On the basis of the previous steps, the last stage is concerned to present a newmodel to evaluate and recognize potential dangerous situations in a level crossingenvironment. This danger evaluation method is built using Hidden Markov Modeland credibility model.Finally, synthetics and real data are used to test the effectiveness and the robustnessof the proposed algorithms and the whole approach by considering various scenarioswithin several situations.This work is developed within the framework of PANsafer project (Towards a saferlevel crossing), supported by the ANR-VTT program (2008) of the French NationalAgency of Research. This project is also labelled by Pôles de compétitivité "i-Trans"and "Véhicule du Futur". All the work, presented in this thesis, has been conductedjointly within IRTES-SET laboratory from UTBM and LEOST laboratory fromIFSTTAR.