Search

To achieve fully autonomous navigation, vehicles need to compute an accurate model of their direct surrounding. In this paper, a 3D surface reconstruction algorithm from heterogeneous density 3D data is presented. The proposed method is based on a TSDF voxel-based representation, where an adaptive neighborhood kernel sourced on a Gaussian confidence evaluation is introduced. This enables to keep a good trade-off between the density of the reconstructed mesh and its accuracy. Experimental evaluations carried on both synthetic (CARLA) and real (KITTI) 3D data show a good performance compared to a state of the art method used for surface reconstruction., IEEE Intelligent Transportation Systems Conference (ITSC) 2019

International audience; Youla-Kucera (YK) parametrization was formulated decades ago for obtaining the set of controllers stabilizing a linear plant. This fundamental result of control theory has been used to develop theoretical tools solving many control problems ranging from stable controller switching, closed-loop identification, robust control, disturbance rejection, adaptive control to fault tolerant control.This paper collects the recent work and classifies the maccording to the use of YK parametrization, Dual YK parametrization or both, providing the latest advances with main applications indifferent control fields. A final discussion gives some insights on the future trends in the field.

The paper proposes a modeling of complex systems to prognose their ability to complete future tasks from local prognostics.The paper presents the decision support indicators.The paper presents inferences to compute the decision support indicators for each node of the model.The general algorithm of the prognostic function is then given.The paper shows how the computed decision support indicators can be used for production and maintenance joint planning. The high costs of complex systems lead companies to improve their efficiency. This improvement can particularly be achieved by reducing their downtimes because of failures or for maintenance purposes. This reduction is the main goal of Condition-Based Maintenance and of Prognostics and Health Management. Both those maintenance policies need to install appropriate sensors and data processes not only to assess the current health of their critical components but also their future health. These future health assessments, also called prognostics, produce the Remaining Useful Life of the components associated to imprecision quantifications. In the case of complex systems where components are numerous, the matter is to assess the health of whole systems from the prognostics of their components (the local prognostics). In this paper, we propose a generic function that assesses the future availability of complex systems from their local prognostics (the prognostics of their components) by using inferences rules. The results of this function can then be used as decision support indicators for planning productive and maintenance tasks. This function exploits a proposed extension for Object Oriented Bayesian Networks (OOBN) used to model the complex system in order to assess the probabilities of failure of components, functions and subsystems. The modeling of the complex system is required and it is presented as well as modeling transformations to tackle some OOBN limitations. Then, the computing inference rules used to define the future availability of complex systems are presented. The extension added to OOBN consists in indicating the components that should first be maintained to improve the availabilities of the functions and subsystems in order to provide a second kind of decision support indicators for maintenance. A fictitious multi-component system bringing together most of the structures encountered in complex systems is modeled and the results obtained from the application of the proposed generic function are presented as well as ways that production and maintenance planning can used the computed indicators. Then we show how the proposed generic prognostic function can be used to predict propagations of failures and their effects on the functioning of functions and subsystems.

We introduce a new approach for multiscale 3Dsemantic scene completion from voxelized sparse 3D LiDAR scans. As opposed to the literature, we use a 2D UNet backbone with comprehensive multiscale skip connections to enhance feature flow, along with 3D segmentation heads. On the SemanticKITTI benchmark, our method performs on par on semantic completion and better on occupancy completion than all other published methods -- while being significantly lighter and faster. As such it provides a great performance/speed trade-off for mobile-robotics applications. The ablation studies demonstrate our method is robust to lower density inputs, and that it enables very high speed semantic completion at the coarsest level. Our code is available at https://github.com/cv-rits/LMSCNet., Accepted at 3DV 2020 (Oral). For a demo video, see http://tiny.cc/lmscnet. Code is available at https://github.com/cv-rits/LMSCNet

The R&D of Autonomous Transportation Systems (ATS) is hindered by the lack of industrial feedback and client's knowledge about technological possibilities. In addition, because of intellectual properties (IP) issues, technology consulting companies can't directly reuse developed functionalities with different clients. In this context, requirements reuse technics presents a good way to capitalize on their knowledge while avoiding IP issues. However, the literature review on requirements reuse processes doesn't propose methods to the application of reuse processes with little information about the system's operational context. In this paper, we present a semi-automated requirement reuse and recycle process for ATS R&D. The process helps designers’ copes with the lack of inputs from the clients. Requirements candidates are retrieved from a database using Natural Language Processing and traceability propagation. It is applied to 3 use cases with inputs less than 5 concepts from the client's needs. The results validate its efficiency through number requirements retrieved and the analysis time consumption

International audience; This paper presents the design and implementation of a novel lateral control approach. The proposed control strategy is based on Youla-Kucera parametrization to switch gradually between controllers that are designed separately for big and small lateral errors. The present approach studies the critical problem of initial lateral error in line following. It ensures smooth and stable transitions between controllers and provides a smooth vehicle response regardless of the lateral error. For an initial validation the present work was tested in simulation, implementing a dynamic bicycle model. It has also been tested in real platforms implementing an electric Renault ZOE, with good results when activating the system at different lateral errors. autonomous vehicles Switched controllers parametrization.

Lidar has become an essential sensor for autonomous driving as it provides reliable depth estimation. Lidar is also the primary sensor used in building 3D maps which can be used even in the case of low-cost systems which do not use Lidar. Computation on Lidar point clouds is intensive as it requires processing of millions of points per second. Additionally there are many subsequent tasks such as clustering, detection, tracking and classification which makes real-time execution challenging. In this paper, we discuss real-time dynamic object detection algorithms which leverages previously mapped Lidar point clouds to reduce processing. The prior 3D maps provide a static background model and we formulate dynamic object detection as a background subtraction problem. Computation and modeling challenges in the mapping and online execution pipeline are described. We propose a rejection cascade architecture to subtract road regions and other 3D regions separately. We implemented an initial version of our proposed algorithm and evaluated the accuracy on CARLA simulator., Comment: Preprint Submission to ECCVW AutoNUE 2018 - v2 author name accent correction

International audience; Perception is one of the most important tasks for safe navigation of autonomous mobile robots in outdoor environments that are dynamic and unknown. This paper presents a real time system for detection and classification of dynamic objects using a 2D laser range finder (2D LRF) and a moving camera mounted on the mobile robot. An occupancy grid based detection is implemented using 2D LRF and moving objects are iteratively extracted from the global occupancy grid. A visual classification system is developed using HOG features and AdaBoost as learning process. Dynamic objects detection and visual classification results are coupled to point moving camera on the specified target (specified by mission supervisor) and to identify it. Finally, real time experiments results on the mobile robot in two different scenarios are presented.

The rapid development of Mobile Internet and Smart Devices and advent of a new generation of Intelligent Transportation Systems (ITS) increase information about present driving conditions and make its prediction possible. Real time traffic information systems (TIS) like SYTADIN help in route to destination planning and traffic state prediction. Energy-optimal routing for electric vehicles creates novel algorithmic challenges where the computation complexity and the quality of information on traffic state are the main issues. This complexity is induced by the possible negative values of edge energy as well as the variability of route and vehicle variables which render the standard algorithms unsuitable. In this paper the authors present an Energy Optimal Real Time Navigation System (EORTNS), implemented on Samsung Galaxy Tab, capable of calculating the route to destination based on information flow obtained from SYTADIN. As an application example the authors propose a real time energy management for a Hybrid Electrical Vehicle (HEV) composed of batteries and Super-Capacitors (SC). The EORTNS is not only capable of energy optimal route to destination calculation with respect to traffic state but also operates the On-Board power splitting between batteries and Super-Capacitors.