A5045448923- Robotics and Sensor-Based Localization
- Indoor and Outdoor Localization Technologies
- Remote Sensing and LiDAR Applications
- Robotic Path Planning Algorithms
- Autonomous Vehicle Technology and Safety
- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- Advanced Vision and Imaging
- Augmented Reality Applications
- Target Tracking and Data Fusion in Sensor Networks
- Underwater Vehicles and Communication Systems
- Inertial Sensor and Navigation
- Air Quality and Health Impacts
- 3D Surveying and Cultural Heritage
- Image and Object Detection Techniques
- Robotic Mechanisms and Dynamics
- Medical Image Segmentation Techniques
- Climate Change and Health Impacts
- Prosthetics and Rehabilitation Robotics
- Urban Heat Island Mitigation
- Data Visualization and Analytics
- Image Processing Techniques and Applications
- Atmospheric chemistry and aerosols
- Mobile Agent-Based Network Management
- Constraint Satisfaction and Optimization
Centre de Robotique
2015-2024
Hôpital Saint-Michel
2015-2024
Airparif
2024
École Nationale Supérieure des Mines de Paris
2018-2023
Centre de Robotique Intégrée d'Ile de France
2019-2021
ParisTech
2016-2020
Université Paris Sciences et Lettres
2016-2020
Research Centre Inria Sophia Antipolis - Méditerranée
2011
Institut national de recherche en informatique et en automatique
2009-2011
Georgia Institute of Technology
2011
ou non, émanant des établissements d'enseignement et de recherche français étrangers, laboratoires publics privés.
This paper presents a metric global localization in the urban environment only with monocular camera and Google Street View database. We fully leverage abundant sources from benefits its topo-metric structure to build coarse-to-fine positioning, namely topological place recognition process then pose estimation by local bundle adjustment. Our method is tested on 3 km demonstrates both sub-meter accuracy robustness viewpoint changes, illumination occlusion. To our knowledge, this first work...
This paper presents an end-to-end real-time monocular absolute localization approach that uses Google Street View panoramas as a prior source of information to train Convolutional Neural Network (CNN). We propose adaptation the PoseNet architecture [8] sparse database panoramas. show we can expand latter by synthesizing new images and consequently improve accuracy pose regressor. The main advantage our method is it does not require first passage equipped vehicle build map. Moreover, offline...
We propose in this letter a tightly-coupled fusion of visual, inertial and magnetic data for long-term localization indoor environment. Unlike state-of-the-art Visual-Inertial SLAM (VISLAM) solutions that reuse visual map to prevent drift, we present an extension the Multi-State Constraint Kalman Filter (MSCKF) takes advantage map. It makes our solution more robust variations environment appearance. The experimental results demonstrate accuracy proposed approach is almost same over time...
This paper introduces a topological localization algorithm that uses visual and Wi-Fi data. Its main contribution is novel way of merging data from these sensors. By making signature suited to FABMAP algorithm, it develops an early-fusion framework solves global kidnapped robot problem. The resulting tested compared localization, over acquired by Pepper in office building. Several constraints were applied during acquisition make the experiment fitted real-life scenarios. Without any tuning,...
Cameras and 2D laser scanners, in combination, are able to provide low-cost, light-weight accurate solutions, which make their fusion well-suited for many robot navigation tasks. However, correct data depends on precise calibration of the rigid body transform between sensors. In this paper we present first framework that makes use Convolutional Neural Networks (CNNs) odometry estimation fusing scanners mono-cameras. The CNNs provides tools not only extract features from two sensors, but also...
The use of 2D laser scanners is attractive for the autonomous driving industry because its accuracy, light-weight and low-cost. However, since only a slice surrounding environment detected at each scan, it challenge to execute important tasks such as localization vehicle. In this paper we present novel framework that explores deep Recurrent Convolutional Neural Networks (RCNN) odometry estimation using scanners. application RCNNs provides tools not extract features scanner data (CNNs), but...
With the fast development of Geographic Information Systems, visual global localization has gained a lot attention due to low price camera and practical implications. In this paper, we leverage Google Street View monocular develop refined continuous positioning in urban environments: namely topological place recognition then 6 DoF pose estimation by local bundle adjustment. order avoid discrete problems, augmented data are virtually synthesized render smooth metric localization. We also...
Automation driving techniques have seen tremendous progresses these last years, particularly due to a better perception of the environment. In order provide safe yet not too conservative in complex urban environment, data fusion should only consider redundant sensing characterize surrounding obstacles, but also be able describe uncertainties and errors beyond presence/absence (be it binary or probabilistic). This paper introduces an enriched representation world, more precisely potential...
This work introduces a new complete Simultaneous Localization and Mapping (SLAM) framework that uses an enriched representation of the world based on sensor fusion is able to simultaneously provide accurate localization vehicle. A method create Evidential grid from two very different sensors, laser scanner stereo camera, allows better handling dynamic aspects urban environment proper management errors more reliable map, thus having precise localization. life-long layer with high level states...
International research is very active in the topic of data fusion between GNSS and proprioceptive sensors to improve basic performances for advanced location-based aiding systems. In this frame, recursive Bayesian estimation methods, still are most efficient popular tools measurement fusion. This paper present comparisons, on one hand two forms Kalman Filter: so-called Linearized Filter (LKF), Extended (EKF), other its promising challengers: Particle (PF). Experimental tests performed...
This article introduces an indoor topological localization algorithm that uses vision and Wi-Fi signals. Its main contribution is a novel way of merging data from these sensors. The designed system does not require knowledge the building plan or positions access points. By making signature suited to FABMAP algorithm, this work develops early fusion framework solves global kidnapped robot problems. resulting has been tested compared with visual localization, over acquired by Pepper in three...
INRIA Sophia Antipolis M´editerran´ee, 2004 route des Lucioles, Cedex, Francecyril.joly@sophia.inria.fr, patrick.rives@sophia.inria.frKeywords: Simultaneous Localization and Mapping (SLAM), Smoothing (SAM), ExtendedKalman Filter (EKF), Bearing-Only, Inverse Depth RepresentationAbstract: Safe efficient navigation in large-scale unknown environments remains a key problem whichhas to be solved improve the autonomy of mobile robots. SLAM methods can bring mapof world trajectory robot. Monucular is...
This paper proposes a new process for calibrating magnetometer in indoor environments that is accurate, easily deployable and time efficient. Our approach simultaneously estimates the calibration of with local variations magnetic field modeled by Gaussian Process. To guarantee an accurate estimation calibration, two-step optimization algorithm proposed. Experiments show proposed solution as outdoor algorithms, more precise than state-of-the-art methods.
The self-localization of a vehicle is still an ongoing and challenging task for the autonomous driving development. At same time, correct understanding vehicle's surroundings creation map with traversed trajectory essential in complex urban scenarios. This paper proposes solution to create enriched global environment while localizing within it. We use Evidential framework based on Dempster-Shafer theory able distinguish between static dynamic obstacles keep information from entire path....
In rough environments, such as off-road or post-crisis drivers often need assistance in piloting their vehicles, especially to anticipate obstacles on the driving path. This research aimed develop a system, focusing cheap and simple method for three-dimensional (3D) reconstruction. step is important detection classification of negative (under-the-ground) positive (above-the-ground) obstacles. information can then be exploited give feedback driver, hence achieving goal driver assistance....
Within a perception framework for autonomous mobile and robotic systems, semantic analysis of 3D point clouds typically generated by LiDARs is key to numerous applications, such as object detection recognition, scene reconstruction. Scene segmentation can be achieved directly integrating spatial data with specialized deep neural networks. Although this type provides rich geometric information regarding the surrounding environment, it also presents challenges: its unstructured sparse nature,...
Autonomous and safe robot navigation requires the capability to simultaneously building a map of environment selflocalization itself. This is known as SLAM (Simultaneous Localization Mapping) problem. In such context, omnidirectional camera looks like very interesting sensor since it allows full 360 degrees field vision. Complexity methods dramatically increases when size grows up. Conversely, accuracy integrity estimation process cannot be guaranteed any more. this paper, we present an...