Papers by Benazouz Bradai
Multiple vehicles based new landmark feature mapping for highly autonomous driving map
A highly autonomous driving (HAD) map can improve the perception and localization performance of ... more A highly autonomous driving (HAD) map can improve the perception and localization performance of autonomous cars. However, when the HAD map cannot represent the real world precisely as a result of changes in the environment, reliability of autonomous cars may be decreased; therefore, it is essential that up-to-date map information is maintained. In order to keep the HAD map up-to-date, new landmark features must be updated continuously. This paper focuses on new landmark feature mapping in the HAD map based on multiple cars equipped with low-cost sensors. The features can be accurately mapped in the HAD map by matching between sensor information and the HAD map information, and by integrating multiple features estimated from various cars. The proposed algorithm is composed of three steps: data acquisition, new landmark feature map estimation using the HAD map, and feature integration. The algorithm is evaluated by means of certain simulation scenarios and experiments.
Entry Mated©: An automated vehicle traffic jam assist function for three pedals entry cars
Optimisation des Lois de Commande d’Éclairage Automobile par Fusion de Données
La conduite nocturne avec un systeme d’eclairage conventionnel n’est pas suffisamment securisante... more La conduite nocturne avec un systeme d’eclairage conventionnel n’est pas suffisamment securisante. En effet, si l’on roule beaucoup moins de nuit, plus de la moitie des accidents de la route arrivent pendant cette periode. Dans le but de reduire ces accidents du trafic nocturne, le projet europeen « Adaptive Front lighting System » (AFS) a ete initie. L’objectif pour les constructeurs et equipementiers automobiles est d’aboutir en 2008 a un changement de reglementation de l’eclairage automobile. Pour cela, ils explorent les differents modes de realisation possible de nouvelles fonctions d’eclairage basees sur la deformation du faisceau lumineux, et etudient la pertinence, l’efficacite par rapport a la situation de conduite, mais aussi les dangers associes a l’utilisation, pour ces fonctions, d’eclairage, d’informations issues du vehicule ou de l’environnement. Depuis 2003, des vehicules proposent d’orienter l’eclairage en virage, cette solution, ne tenant compte que des actions du conducteur sur le volant, permet d’ameliorer la visibilite en orientant le faisceau vers l’interieur du virage. Cependant, le profil de route (intersections, courbure, etc. ) n’etant pas toujours connu du conducteur, les performances liees a cette solution sont des lors limitees. Or les systemes embarques de navigation, d’une part peuvent apporter des informations primordiales sur cette forme, et d’autre part disposent d’informations contextuelles (ouvrages d’art, nature de la route, rayon de virage, limitations de vitesse en vigueur…). Le sujet de cette these a pour objectif d’optimiser les lois de commande des systemes d’eclairage en s’appuyant sur la fusion des informations issues des systemes de navigation avec celles des capteurs embarques dans le vehicule (cameras, …), tout en sachant jusqu’a quel point les systemes actuels et futurs peuvent repondre a ces attentes de maniere efficace et fiable. Ainsi, cette fusion des informations, appliquee ici a la prise de decision, permet de definir les situations et les contextes de conduite de l’environnement d’evolution du vehicule (autoroute, ville, etc. ) et de choisir la loi appropriee parmi les differentes lois de commande d’eclairage developpees pour repondre aux fonctionnalites recherchees (code autoroute, code ville, code virage). Cette demarche permet de choisir en temps reel, et par anticipation, entre ces differentes lois de commande. Elle permet, par consequent, l’amelioration de la robustesse du systeme d’eclairage. Deux points sont a l’origine de cette amelioration. Premierement, a partir du systeme de navigation, nous avons developpe un capteur virtuel d’horizon glissant evenementiel permettant la determination precise des differentes situations de conduite en utilisant un automate d’etats finis. Il permet ainsi de pallier aux problemes de la nature ponctuelle des informations du systeme de navigation. Deuxiemement, nous avons developpe un capteur virtuel generique de determination des situations de conduite base sur la theorie des croyances en utilisant un systeme de navigation et la vision. Ce capteur combine les confiances en provenance des deux sources pour mieux distinguer les differentes situations et les differents contextes de conduite et de pallier aux problemes des deux sources prises independamment. Il permet egalement de construire une confiance du systeme de navigation. Ce capteur generique est generalisable a des systemes d’aide a la conduite (ADAS) autre que l’eclairage. Ceci a ete montre en l’appliquant a un systeme de detection des limitations de vitesses reglementaires SLS (Speed Limit Support). Les deux capteurs virtuels developpes ont ete appliques a l’optimisation de l’eclairage AFS et au systeme SLS. Ces deux systemes ont ete implementes sur un vehicule de demonstration et ils sont actuellement operationnels. Ils ont ete evalues par differents types de conducteur allant des non experts aux experts de l’eclairage et des systemes d’aide a la conduite (ADAS). Ils ont ete egalement montres aupres des constructeurs automobiles (PSA, Audi, Renault, Honda, etc. ) et au cours de differents « techdays » et ils ont prouve leur fiabilite lors des demonstrations sur routes ouvertes avec des differentes situations et differents contextes de conduite.
Loosely-coupled localization fusion system based on track-to-track fusion with bias alignment
IEEE Access
In autonomous vehicles, perception information about the surrounding road environment can be obta... more In autonomous vehicles, perception information about the surrounding road environment can be obtained through image semantic segmentation. The fisheye camera commonly used in autonomous vehicle surround view systems offers a wide field of view (FoV), providing comprehensive perception information about the surrounding environment and assisting in understanding complex scenes. However, there is a challenge in model training due to the limited availability of fisheye semantic image datasets, resulting in reduced generalization performance and unreliable results in various test environments. In particular, changes in the position and orientation of the camera result in changes in the camera viewpoint, which can impair the model's segmentation performance. Generally, data scarcity problems are solved using augmentation methods, but existing methods have difficulty reflecting the distortion characteristics of fisheye images. To solve this problem, we propose viewpoint augmentation considering the spatially variant distortion characteristic of fisheye images. First, we use the fisheye camera projection model in reverse to map the captured 2D fisheye image to a point on the surface of a unit sphere in 3D. Then, we change the camera's orientation and position by applying rotation and translation operations to the point. Finally, we reproject the transformed point to the fisheye image to generate a fisheye image with a changed viewpoint. The experimental results show that the proposed augmentation method increases the generalization performance of the model and effectively reduces model performance degradation under changing camera viewpoints, making it suitable for practical applications. INDEX TERMS Camera viewpoint change, data augmentation, surround-view fisheye camera, image semantic segmentation, intelligent vehicles.
Journal of Advanced Transportation
LiDAR-based localization has been widely used for the pose estimation of autonomous vehicles. Sin... more LiDAR-based localization has been widely used for the pose estimation of autonomous vehicles. Since the localization requires a sustainable map reflecting environment changes, a map update framework based on crowd-sourcing measurements has been researched. Unfortunately, a point cloud map occupies too large data size to transmit data in the uploading and downloading of the map update framework. To realize the LiDAR map update framework by reducing the data size, we proposed a novel map update framework using a Geodetic Normal Distribution (GND) map that compresses the point cloud to the normal distributions. The proposed GND map update framework comprises two parts: map change detection based on crowd-sourcing vehicles and map updating based on a map cloud server. GND map changes are detected based on an evidence theory considering geometric relationships between the GND map and crowd-sourcing measurements and uploaded to the map cloud server. Uploaded map changes reproduce represen...
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific ... more HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et a ̀ la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Tracking both pose and status of a traffic light via
Detection of traffic lights from image
The invention relates to a method for detecting the presence and the state of a traffic light on ... more The invention relates to a method for detecting the presence and the state of a traffic light on a road scene from the processing of images captured by at least one camera mounted on a motor vehicle. According to the invention, one traffic light is detected and monitoring (S2, S4, S5) from the captured images (If) by a first front camera and from captured images (S3) by at least one camera vision of a plurality of vision cameras on-board peripheral vision system. Application to the management of automatic stops and starts of an autonomous vehicle based on the state of a traffic light.
Beam Control Method for Lighting of Vehicles
The invention relates to a method and an illumination beam for vehicle control system. It propose... more The invention relates to a method and an illumination beam for vehicle control system. It proposes controlling the road lighting beam by combining data obtained from a navigation device and detection means adapted to provide distance information and such as to target objects (7) located in front of the vehicle ( 1) equipped.
A method for automatically determining the speed limit on a line segment and associated system
Method for automatic determination of speed limits on the road system and associated
The present invention essentially relates to an automatic method for determining a current speed ... more The present invention essentially relates to an automatic method for determining a current speed limit on a road traveled by a motor vehicle, wherein: - is established by means of a first system (101), involving in particular a GPS antenna and mapping data, a first set of information (151) combining at least one probable speed limit of a confidence index (IC1); - is established by means of a second system (102), including through a camera and image processing applications capable of identifying speed limit signs arranged in the vicinity of the road, a second set of information comprising at least one probable speed limit; - determining (154) from the first set of information and the second set of information, the current speed limit on the road considered.
Multiple vehicles based new landmark feature mapping for highly autonomous driving map
2017 14th Workshop on Positioning, Navigation and Communications (WPNC), 2017
A highly autonomous driving (HAD) map can improve the perception and localization performance of ... more A highly autonomous driving (HAD) map can improve the perception and localization performance of autonomous cars. However, when the HAD map cannot represent the real world precisely as a result of changes in the environment, reliability of autonomous cars may be decreased; therefore, it is essential that up-to-date map information is maintained. In order to keep the HAD map up-to-date, new landmark features must be updated continuously. This paper focuses on new landmark feature mapping in the HAD map based on multiple cars equipped with low-cost sensors. The features can be accurately mapped in the HAD map by matching between sensor information and the HAD map information, and by integrating multiple features estimated from various cars. The proposed algorithm is composed of three steps: data acquisition, new landmark feature map estimation using the HAD map, and feature integration. The algorithm is evaluated by means of certain simulation scenarios and experiments.
Method of early detection of a bend on a portion of road and associated system
Method for Determining, in a Predictive Manner, Types of Road Situations for a Vehicle
Method for the anticipated ascertainment of a bend on a portion of road, and associated system
Method for Controlling a Vehicle Member
Process for the Automatic Determination of Speed Limitations on a Road and an Associated System
Optimisation des Lois de Commande d’Éclairage Automobile par Fusion de Données
Http Www Theses Fr, 2007
La conduite nocturne avec un systeme d’eclairage conventionnel n’est pas suffisamment securisante... more La conduite nocturne avec un systeme d’eclairage conventionnel n’est pas suffisamment securisante. En effet, si l’on roule beaucoup moins de nuit, plus de la moitie des accidents de la route arrivent pendant cette periode. Dans le but de reduire ces accidents du trafic nocturne, le projet europeen « Adaptive Front lighting System » (AFS) a ete initie. L’objectif pour les constructeurs et equipementiers automobiles est d’aboutir en 2008 a un changement de reglementation de l’eclairage automobile. Pour cela, ils explorent les differents modes de realisation possible de nouvelles fonctions d’eclairage basees sur la deformation du faisceau lumineux, et etudient la pertinence, l’efficacite par rapport a la situation de conduite, mais aussi les dangers associes a l’utilisation, pour ces fonctions, d’eclairage, d’informations issues du vehicule ou de l’environnement. Depuis 2003, des vehicules proposent d’orienter l’eclairage en virage, cette solution, ne tenant compte que des actions du conducteur sur le volant, permet d’ameliorer la visibilite en orientant le faisceau vers l’interieur du virage. Cependant, le profil de route (intersections, courbure, etc. ) n’etant pas toujours connu du conducteur, les performances liees a cette solution sont des lors limitees. Or les systemes embarques de navigation, d’une part peuvent apporter des informations primordiales sur cette forme, et d’autre part disposent d’informations contextuelles (ouvrages d’art, nature de la route, rayon de virage, limitations de vitesse en vigueur…). Le sujet de cette these a pour objectif d’optimiser les lois de commande des systemes d’eclairage en s’appuyant sur la fusion des informations issues des systemes de navigation avec celles des capteurs embarques dans le vehicule (cameras, …), tout en sachant jusqu’a quel point les systemes actuels et futurs peuvent repondre a ces attentes de maniere efficace et fiable. Ainsi, cette fusion des informations, appliquee ici a la prise de decision, permet de definir les situations et les contextes de conduite de l’environnement d’evolution du vehicule (autoroute, ville, etc. ) et de choisir la loi appropriee parmi les differentes lois de commande d’eclairage developpees pour repondre aux fonctionnalites recherchees (code autoroute, code ville, code virage). Cette demarche permet de choisir en temps reel, et par anticipation, entre ces differentes lois de commande. Elle permet, par consequent, l’amelioration de la robustesse du systeme d’eclairage. Deux points sont a l’origine de cette amelioration. Premierement, a partir du systeme de navigation, nous avons developpe un capteur virtuel d’horizon glissant evenementiel permettant la determination precise des differentes situations de conduite en utilisant un automate d’etats finis. Il permet ainsi de pallier aux problemes de la nature ponctuelle des informations du systeme de navigation. Deuxiemement, nous avons developpe un capteur virtuel generique de determination des situations de conduite base sur la theorie des croyances en utilisant un systeme de navigation et la vision. Ce capteur combine les confiances en provenance des deux sources pour mieux distinguer les differentes situations et les differents contextes de conduite et de pallier aux problemes des deux sources prises independamment. Il permet egalement de construire une confiance du systeme de navigation. Ce capteur generique est generalisable a des systemes d’aide a la conduite (ADAS) autre que l’eclairage. Ceci a ete montre en l’appliquant a un systeme de detection des limitations de vitesses reglementaires SLS (Speed Limit Support). Les deux capteurs virtuels developpes ont ete appliques a l’optimisation de l’eclairage AFS et au systeme SLS. Ces deux systemes ont ete implementes sur un vehicule de demonstration et ils sont actuellement operationnels. Ils ont ete evalues par differents types de conducteur allant des non experts aux experts de l’eclairage et des systemes d’aide a la conduite (ADAS). Ils ont ete egalement montres aupres des constructeurs automobiles (PSA, Audi, Renault, Honda, etc. ) et au cours de differents « techdays » et ils ont prouve leur fiabilite lors des demonstrations sur routes ouvertes avec des differentes situations et differents contextes de conduite.
IEEE Access
In the autonomous car, perception with point cloud semantic segmentation helps obtain a wealth of... more In the autonomous car, perception with point cloud semantic segmentation helps obtain a wealth of information about the surrounding road environment. Despite the massive progress of recent researches, the existing machine learning networks are still insufficient for online applications of autonomous driving due to too subdivided classes, the lack of training data, and their heavy computing load. To solve these problems, we propose a fast and lite point cloud semantic segmentation network for autonomous driving, which utilizes LiDAR synthetic data to improve the performance by transfer learning. First, we modify the labeling classes and generate the LiDAR synthetic data-set for additional training to alleviate the lack of training data of the realistic data-set. Then, to lower the computing load, we adopt a projection-based method and apply a lightweight segmentation network to projected LiDAR images, which has drastically reduced computing. Finally, we verified and evaluated the proposed network in this paper through experiments. Experimental results show that the proposed network can perform the three-dimensional point cloud semantic segmentation in an online way, in which the inference speed overwhelms the existing algorithms. INDEX TERMS Autonomous driving, point cloud semantic segmentation, synthetic dataset, computing load reduction.
IEEE Access, 2021
Light detection and ranging (LiDAR) sensors enable a vehicle to estimate a pose by matching their... more Light detection and ranging (LiDAR) sensors enable a vehicle to estimate a pose by matching their measurements with a point cloud (PCD) map. However, the PCD map structure, widely used in robot fields, has some problems to be applied for mass production in automotive fields. First, the PCD map is too big to store all map data at in-vehicle units or download the map data from a wireless network according to the vehicle location. Second, the PCD map, represented by a single origin in the Cartesian coordinates, causes coordinate conversion errors due to an inaccurate plane-orb projection, when the vehicle estimate the geodetic pose on Earth. To solve two problems, this paper presents a geodetic normal distribution (GND) map structure. The GND map structure supports a geodetic quad-tree tiling system with multiple origins to minimize the coordinate conversion errors. The map data managed by the GND map structure are compressed by using Cartesian probabilistic distributions of points as map features. The truncation errors by heterogeneous coordinates between the geodetic tiling system and Cartesian distributions are compensated by the Cartesian voxelization rule. In order to match the LiDAR measurements with the GND map structure, the paper proposes map-matching approaches based on Monte-Carlo and optimization. The paper performed some experiments to evaluate the map size compression and the long-term localization on Earth: comparison with the PCD map structure, localization in various continents, and long-term localization. INDEX TERMS World-scale map management, map compression, normal distribution map, registration.
Uploads
Papers by Benazouz Bradai