Papers by Steve Pechberti
Integrated neuro-robotic approaches for autonomous vehicle localisation and navigation
Decentralized Perception System with Multiple Viewpoints
SAE Technical Paper Series
Vehicle-to-Infrastructure (V2I) cooperation has emerged as a fundamental technology to overcome t... more Vehicle-to-Infrastructure (V2I) cooperation has emerged as a fundamental technology to overcome the limitations of the individual ego-vehicle perception. Onboard perception is limited by the lack of information for understanding the environment, the lack of anticipation, the drop of performance due to occlusions and the physical limitations of embedded sensors. The perception of V2I in a cooperative manner improves the perception range of the ego vehicle by receiving information from the infrastructure that has another point of view, mounted with sensors, such as camera and LiDAR. This technical paper presents a perception pipeline developed for the infrastructure based on images with multiple viewpoints. It is designed to be scalable and has five main components: the image acquisition for the modification of camera settings and to get the pixel data, the object detection for fast and accurate detection of four wheels, two wheels and pedestrians, the data fusion module for robust fu...
Lecture Notes in Computer Science, 2022
Efficient encoding of visual information is essential to the success of vision-based navigation t... more Efficient encoding of visual information is essential to the success of vision-based navigation tasks in large-scale environments. To do so, we propose in this article the Sparse Max-Pi neural network (SMP), a novel compute-efficient model of visual localization based on sparse and topological encoding of visual information. Inspired by the spatial cognition of mammals, the model uses a "topologic sparse dictionary" to efficiently compress the visual information of a landmark, allowing rich visual information to be represented with very small codes. This descriptor, inspired by the neurons in the primary visual cortex (V1), are learned using sparse coding, homeostasis and self-organising map mechanisms. Evaluated in cross-validation on the Oxford-car dataset, our experimental results show that the SMP model is competitive with the state of the art. It thus provides comparable or better performance than CoHog and NetVlad, two state-of-the-art VPR models.
Lecture Notes in Computer Science, 2022
Efficient encoding of visual information is essential to the success of vision-based navigation t... more Efficient encoding of visual information is essential to the success of vision-based navigation tasks in large-scale environments. To do so, we propose in this article the Sparse Max-Pi neural network (SMP), a novel compute-efficient model of visual localization based on sparse and topological encoding of visual information. Inspired by the spatial cognition of mammals, the model uses a "topologic sparse dictionary" to efficiently compress the visual information of a landmark, allowing rich visual information to be represented with very small codes. This descriptor, inspired by the neurons in the primary visual cortex (V1), are learned using sparse coding, homeostasis and self-organising map mechanisms. Evaluated in cross-validation on the Oxford-car dataset, our experimental results show that the SMP model is competitive with the state of the art. It thus provides comparable or better performance than CoHog and NetVlad, two state-of-the-art VPR models.
ArXiv, 2022
Pedestrian crossing prediction has been a topic of active research, resulting in many new algorit... more Pedestrian crossing prediction has been a topic of active research, resulting in many new algorithmic solutions. While measuring the overall progress of those solutions over time tends to be more and more established due to the new publicly available benchmark and standardized evaluation procedures, knowing how well existing predictors react to unseen data remains an unanswered question. This evaluation is imperative as serviceable crossing behavior predictors should be set to work in various scenarii without compromising pedestrian safety due to misprediction. To this end, we conduct a study based on direct cross-dataset evaluation. Our experiments show that current state-of-the-art pedestrian behavior predictors generalize poorly in cross-dataset evaluation scenarii, regardless of their robustness during a direct training-test set evaluation setting. In the light of what we observe, we argue that the future of pedestrian crossing prediction, e.g. reliable and generalizable impleme...
2010 IEEE Intelligent Vehicles Symposium, 2010
This paper describes hybrid fusion module used in a strong localization context (POMA) for embedd... more This paper describes hybrid fusion module used in a strong localization context (POMA) for embedded vehicle applications. This work has been developed in order to give an answer to the POMA (Positioning, Maps and local referencing) sub project objectives. These objectives are to provide, for a set of high level applications, a positioning service included a service quality, a metric accuracy (lane) and a robust result. This work is involved in CVIS European project. The use of IMM approach in the Hybrid Fusion module will be justified in comparison to the different current probabilistic methods. The IMM, contrary to the non modular methods, is based on the discretization of the vehicle evolution space into simple maneuvers, represented each by a simple dynamic model such as constant velocity or constant turning etc. This allows the method to be optimized for highly dynamic vehicles. The application of this positioning service will be presented in a real time embedded architecture. The presented results are based on real measurements collected from representatives scenarios (test track, peri-urban road, highway). These results show a real interest in using the new IMM method in order to reach the POMA's objectives.
Proposal of a virtual and immersive 3D architecture dedicated for prototyping, test and evaluation of eco-driving applications
2013 IEEE Intelligent Vehicles Symposium (IV), 2013
ABSTRACT Simulation has been widely used to estimate the benefits of ADAS with embedded sensors o... more ABSTRACT Simulation has been widely used to estimate the benefits of ADAS with embedded sensors or more recently Cooperative Systems based on Inter-Vehicular Communications. This paper presents the proposal of a new architecture built with the both SiVIC and RTMaps platforms in order to prototype, to test and to validate eco-driving applications. In this architecture, the innovation is mainly due to the real-time immersion of a real driver in a 3D virtual environment. In this “Hardware In the Loop” platform, major contributions have been made about vehicle modeling updating, interconnection between SiVIC and an HMD. Moreover, a first modeling of a consumption sensor is proposed and used in order to achieve some tests of fuel consumption on the virtual Satory's track. All these contributions have been tuned with on-road measurements to improve reality of the scenarios. We discuss the results of a simple eco-driving scenario implemented to validate our architecture's capabilities.
2013 IEEE Intelligent Vehicles Symposium Workshops (IV Workshops), 2013
LIVIC-IFSTTAR develops driving assistance services in order to improve the driving safety. These ... more LIVIC-IFSTTAR develops driving assistance services in order to improve the driving safety. These systems are tested on several real prototypes equipped with sensors and perception, decision and control modules. But tests on real prototypes are not always available, effectively some hardware architectures could be too expensive to implement, scenario may lead to hazardous situations. Moreover, lots of reasons could lead to the inability to obtain both sensors and ground truth data for ADAS evaluation. However, safety applications must be tested in order to guaranty their reliability. For this task, simulation appears as a good alternative to the real prototyping and testing stages. In this context, the simulation must provide the same opportunities as reality, by providing all the necessary data to develop and to prototype different types of ADAS based on local or extended environment perception. The sensor data provided by simulation must be as noised and imperfect as those obtained with real sensors. To address this issue, the SiVIC platform has been developed; it provides a virtual road environment including realistic dynamic models of mobile entities (vehicles), realistic sensors, and sensors for ground truth. To test real embedded applications, an interconnection has been developed between SiVIC and third party applications (ie. RTMaps). In this way, the prototyped application can be directly embedded in real prototypes in order to test it in real conditions. A Full Speed Range ACC application is presented in this paper to illustrate the capabilities and the functionalities of this virtual platform.
2012 15th International IEEE Conference on Intelligent Transportation Systems, 2012
This paper proposes a new radar sensor modelling for Advanced Driver Assistance Systems (ADAS) pr... more This paper proposes a new radar sensor modelling for Advanced Driver Assistance Systems (ADAS) prototyping. The model is embedded on the SiVIC platform (Simulator for Vehicle, Infrastructure and Sensors). Lots of simulators already exist for this issue, but none is designed to address the objectives of real-time computation, highly sampled signal generation. And few simulators offer the ability to be integrated in a dynamic platform for the ADAS prototyping. In this paper, several radar technologies will be presented. Then, a radar designed especially for automotive domain will be described exploring each subparts, radar antenna andi.e. propagation channel. Such as the generic model, hypothesis done on electromagnetic waves and environmental objects modelling will also be provided. A first model of simple duplex radar with Frequency Shift Keying (FSK) modulation is implemented and shown as illustration for the defined architecture. Finally, in order to optimize the duration for signal generation, several software architecture solution will be proposed.
2011 IEEE Intelligent Vehicles Symposium (IV), 2011
Inter-Vehicular Communications (IVC) are considered a promising technological approach for enhanc... more Inter-Vehicular Communications (IVC) are considered a promising technological approach for enhancing transportation safety and improving highway efficiency. Previous theoretical work has demonstrated the benefits of IVC in vehicles strings. Simulations of partially IVC-equipped vehicles strings showed that only a small equipment ratio is sufficient to drastically reduce the number of head on collisions. However, these results are based on the assumptions that IVC exhibit lossless and instantaneous messages transmission. This paper presents the research design of an empirical measurement of a vehicles string, with the goal of highlighting the constraints introduced by the actual characteristics of communication devices. A warning message diffusion system based on IEEE 802.11 wireless technology was developed for an emergency breaking scenario. Preliminary results are presented as well, showing the latencies introduced by using 802.11a and discussing early findings and experimental limitations
Optimized simulation architecture for multimodal radar modeling: Application to automotive driving assistance system
16th International IEEE Conference on Intelligent Transportation Systems (ITSC 2013), 2013
ABSTRACT This paper presents a new radar sensor modeling for Advanced Driver Assistance Systems (... more ABSTRACT This paper presents a new radar sensor modeling for Advanced Driver Assistance Systems (ADAS) prototyping. This virtual model is embedded in the SiVIC platform dedicated to the simulation of vehicles, infrastructure and sensors. Lots of simulators already exist addressing radar simulation, but none of them is designed to address in the same time the following objectives: real-time computation, highly sampled signal generation, multiple radar technologies, and, hardware and software platform integration for ADAS prototyping. In this paper, we proposed such a solution solving this issue and a focus will be done on the radar sensor modeling, from material modeling to implemented technology and associated signal processing. First results of this radar sensor modeling is carried out in motorway scenario.
It now seems essential to take into account the information concerning the distant road environme... more It now seems essential to take into account the information concerning the distant road environment. This extended perception is necessary to ensure an efficient capacity to predict events and thus to react appropriately. Moreover, information should not be managed only on surrounding moving objects (local perception) but also on both the configuration of the distant road infrastructure and the weather conditions in which the different actors of the road system evolve. This implies to develop and to implement of cooperative systems combining embedded processing, processing on the infrastructure (road side unit) and communication media to make the link between the different information sources. In this paper, this type of cooperative application is presented in the framework of the DIVAS's project. Moreover, a dedicated distributed virtual architecture is proposed in order to prototype, to test and to validate such type of cooperative application.
2021 16th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2021)
Understanding the behaviors and intentions of pedestrians is still one of the main challenges for... more Understanding the behaviors and intentions of pedestrians is still one of the main challenges for vehicle autonomy, as accurate predictions of their intentions can guarantee their safety and driving comfort of vehicles. In this paper, we address pedestrian crossing prediction in urban traffic environments by linking the dynamics of a pedestrian's skeleton to a binary crossing intention. We introduce TrouSPI-Net: a context-free, lightweight, multi-branch predictor. TrouSPI-Net extracts spatio-temporal features for different time resolutions by encoding pseudo-images sequences of skeletal joints' positions and processes them with parallel attention modules and atrous convolutions. The proposed approach is then enhanced by processing features such as relative distances of skeletal joints, bounding box positions, or ego-vehicle speed with U-GRUs. Using the newly proposed evaluation procedures for two large public naturalistic data sets for studying pedestrian behavior in traffic: JAAD and PIE, we evaluate TrouSPI-Net and analyze its performance. Experimental results show that TrouSPI-Net achieved 76% F1 score on JAAD and 80% F1 score on PIE, therefore outperforming current state-of-the-art while being lightweight and context-free.
Design of a modular demonstrator for safety application systems: The CVIS project
Abstract—This article describes the design of a modular
Modélisation et simulation de capteurs électromagnétiques appliquées au domaine automobile pour le prototypage de systèmes d'aide à la conduite. Applications aux radars et systèmes de télécommunications
Cette these propose un modele de capteur electromagnetique generique pour la simulation de scenes... more Cette these propose un modele de capteur electromagnetique generique pour la simulation de scenes routieres dans le cadre de la mise en oeuvre de nouveaux systemes d'aide a la conduite pour le transport terrestre. L'une des contraintes majeures de cette problematique concerne les temps de traitements disponibles dans un tel contexte. Pour obtenir ou approcher d'un fonctionnement temps reel, plusieurs methodes de resolution utilisant les capacites de l'architecture centrale (CPU) et les capacites programmables des cartes graphiques (GPU) sont mises en oeuvres. Un bus de partage est egalement defini afin de pouvoir repartir les traitements relatifs aux modeles presents dans la scene virtuelle sur un ensemble de poste de travail. La simulation de la propagation des ondes electromagnetiques est abordee en fonction des situations etudiees et du niveau de granularite souhaite. En general, ce niveau de representation est directement lie aux bruits associes a l'environne...
Understanding the behaviors and intentions of pedestrians is still one of the main challenges for... more Understanding the behaviors and intentions of pedestrians is still one of the main challenges for vehicle autonomy, as accurate predictions of their intentions can guarantee their safety and driving comfort of vehicles. In this paper, we address pedestrian crossing prediction in urban traffic environments by linking the dynamics of a pedestrian’s skeleton to a binary crossing intention. We introduce TrouSPI-Net: a context-free, lightweight, multi-branch predictor. TrouSPI-Net extracts spatio-temporal features for different time resolutions by encoding pseudo-images sequences of skeletal joints’ positions and processes them with parallel attention modules and atrous convolutions. The proposed approach is then enhanced by processing features such as relative distances of skeletal joints, bounding box positions, or ego-vehicle speed with UGRUs. Using the newly proposed evaluation procedures for two large public naturalistic data sets for studying pedestrian behavior in traffic: JAAD an...
Algorithms
Understanding the behaviors and intentions of humans is still one of the main challenges for vehi... more Understanding the behaviors and intentions of humans is still one of the main challenges for vehicle autonomy. More specifically, inferring the intentions and actions of vulnerable actors, namely pedestrians, in complex situations such as urban traffic scenes remains a difficult task and a blocking point towards more automated vehicles. Answering the question “Is the pedestrian going to cross?” is a good starting point in order to advance in the quest to the fifth level of autonomous driving. In this paper, we address the problem of real-time discrete intention prediction of pedestrians in urban traffic environments by linking the dynamics of a pedestrian’s skeleton to an intention. Hence, we propose SPI-Net (Skeleton-based Pedestrian Intention network): a representation-focused multi-branch network combining features from 2D pedestrian body poses for the prediction of pedestrians’ discrete intentions. Experimental results show that SPI-Net achieved 94.4% accuracy in pedestrian cros...
2010 IEEE Intelligent Vehicles Symposium, 2010
This article describes the design of a modular demonstrator (Enhanced Driver Awarness-EDA) aimed ... more This article describes the design of a modular demonstrator (Enhanced Driver Awarness-EDA) aimed at integrating various safety embedded vehicular applications. Two goals are achieved with this demonstrator: the first one is to provide a complete set of useful function in order to develop safety applications; the other one is to allow multiple services to run simultaneously on the same vehicle sharing information to improve their potentials. This demonstrator has been used in the European CVIS project, which aim is to provide a set of technologies defined for embedded safety vehicle application.
2010 IEEE Intelligent Vehicles Symposium, 2010
This article describes the design of a modular demonstrator (Enhanced Driver Awarness-EDA) aimed ... more This article describes the design of a modular demonstrator (Enhanced Driver Awarness-EDA) aimed at integrating various safety embedded vehicular applications. Two goals are achieved with this demonstrator: the first one is to provide a complete set of useful function in order to develop safety applications; the other one is to allow multiple services to run simultaneously on the same vehicle sharing information to improve their potentials. This demonstrator has been used in the European CVIS project, which aim is to provide a set of technologies defined for embedded safety vehicle application.
Uploads
Papers by Steve Pechberti