HAL - Lille 3
Not a member yet
5967 research outputs found
Sort by
Virtual 3D city model as a priori information source for vehicle localization system
No full textInternational audienceThis paper aims at demonstrating the usefulness of integrating virtual 3D models in vehicle ego-localization systems. Usually, vehicle localization algorithms are based on multi-sensor data fusion. Global Navigation Satellite Systems GNSS, as Global Positioning System GPS, are used to provide measurements of the geographic location. Nevertheless, GNSS solutions suffer from signal attenuation and masking, multipath phenomena and lack of visibility, especially in urban areas. That leads to degradation or even a total loss of the positioning information and then unsatisfactory performances. Dead-reckoning and inertial sensors are then often added to back up GPS in case of inaccurate or unavailable measurements or if high frequency location estimation is required. However, the dead-reckoning localization may drift in the long term due to error accumulation. To back up GPS and compensate the drift of the dead reckoning sensors based localisation, two approaches integrating a virtual 3D model are proposed in sregistered with respect to the scene perceived by an on-board sensor. From the real/virtual scenes matching, the transformation (rotation and translation) between the real sensor and the virtual sensor (whose position and orientation are known) can be computed. These two approaches lead to determine the pose of the real sensor embedded on the vehicle. In the first approach, the considered perception sensor is a camera and in the second approach, it is a laser scanner. The first approach is based on image matching between the virtual image extracted from the 3D city model and the real image acquired by the camera. The two major parts are: 1/ Detection and matching of feature points in real and virtual images (three features points are compared: Harris corner detector, SIFT and SURF); 2/ Pose computation using POSIT algorithm. The second approach is based on the on-board horizontal laser scanner that provides a set of distances between it and the environment. This set of distances is matched with depth information (virtual laser scan data), provided by the virtual 3D city model. The pose estimation provided by these two approaches can be integrated in data fusion formalism. In this paper the result of the first approach is integrated in IMM UKF data fusion formalism. Experimental results obtained using real data illustrate the feasibility and the performances of the proposed approaches
Switching Time Estimation and Active Mode Recognition using a Data Projection Method
No full textInternational audienceThis paper proposes a data projection method (DPM) to detect a mode switching and recognize the current mode in a switching system. The main feature of this method is that the precise knowledge of the system model, i.e., the parameter values, is not needed. One direct application of this technique is fault detection and identification (FDI) when a fault produces a change in the system dynamics. Mode detection and recognition correspond to fault detection and identification, and switching time estimation to fault occurrence time estimation The general principle of the DPM is to generate mode indicators, namely, residuals, using matrix projection techniques, where matrices are composed of input and output measured data. The DPM is presented in detail, and properties of switching detectability (fault detectability) and discernability between modes (fault identifiability) are characterized and discussed. The great advantage of this method, compared with other techniques in the literature, is that it does not need the model parameter values and thus can be applied to systems of the same type without identifying their parameters. This is particularly interesting in the design of generic embedded fault diagnosis algorithms
"What Is the City but the People?" Exploring Urban Activity Using Social Web Traces
No full textInternational audienceWe demonstrate GeoTopics, a system to explore geographical patterns of urban activity. The system collects publicly shared check-ins generated by Foursquare users, that reveal who spends time where, when, and on what type of activity. It then employs sparse probabilistic modeling techniques to learn associations between different regions of a city and multi-feature descriptions of urban activity. Through a web interface, users of the system can select a city of interest and explore visualizations that highlight how different types of activity are spatially and temporally distributed in the city. We discuss the opportunities that web data offer to understand urban activity and the challenges one faces in that task. We then describe our approach and the architecture of GeoTopics. Finally, we lay out the demonstration scenario
Designing biomedical proteomics experiments: state-of-the-art and future perspectives
No full textInternational audienceWith the current expanded technical capabilities to perform mass spectrometry-based biomedi-cal proteomics experiments, an improved focus on the design of experiments is crucial. As it is clear that ignoring the importance of a good design leads to an unprecedented rate of false discoveries which would poison our results, more and more tools are developed to help researchers designing proteomic experiments. In this review, we apply statistical thinking to go through the entire proteomics workflow for biomarker discovery and validation and relate the considerations that should be made at the level of hypothesis building, technology selection, experimental design and the optimization of the experimental parameters
Collaborative Localization for Multi-Robot System with Fault Detection and Exclusion based on the Kullback-Leibler Divergence
No full textInternational audienceMulti-robot system attracted attention in various applications in order to replace the human operators. To achieve the intended goal, one of the main challenges of this system is to ensure the integrity of localization by adding a sensor fault diagnosis step to the localization task. In this paper, we present a framework able, in addition of localizing a group of robots, to detect and exclude the faulty sensors from the group with an optimized thresholding method. The estimator has the informational form of the Kalman Filter (KF) namely Information Filter (IF). A residual test based on the Kullback-Leibler divergence (KLD) between the predicted and the corrected distributions of the IF is developed. It is generated from two tests: the first acts on the means and the second deals with the covariance matrices. Thresholding using entropy based criterion and Receiver Operating Characteristics (ROC) curve are discussed. Finally, the validation of this framework is studied on real experimental data from a group of robots
Random Shuffling and Resets for the Non-stationary Stochastic Bandit Problem
No full textWe consider a non-stationary formulation of the stochastic multi-armed bandit where the rewards are no longer assumed to be identically distributed. For the best-arm identification task, we introduce a version of SUCCESSIVE ELIMINATION based on random shuffling of the K arms. We prove that under a novel and mild assumption on the mean gap ∆, this simple but powerful modification achieves the same guarantees in term of sample complexity and cumulative regret than its original version, but in a much wider class of problems, as it is not anymore constrained to stationary distributions. We also show that the original SUCCESSIVE ELIMINATION fails to have controlled regret in this more general scenario, thus showing the benefit of shuffling. We then remove our mild assumption and adapt the algorithm to the best-arm identification task with switching arms. We adapt the definition of the sample complexity for that case and prove that, against an optimal policy with N − 1 switches of the optimal arm, this new algorithm achieves an expected sample complexity of O(∆^{−2} sqrt(N Kdelta^{−1} log(K/delta)), where δ is the probability of failure of the algorithm, and an expected cumulative regret of O(∆^{−1} sqrt(N T K log(T K))) after T time steps
Coordinated predictive control in active distribution networks with HV/MV reactive power constraint
No full textInternational audienceThis paper presents a new real time centralized Model Predictive Control algorithm for distribution networks. Compared to existing works regarding MPC volt var control, the proposed algorithm controls not only the MV voltages but also the reactive power exchange at the distribution and transmission systems interface. Control of reactive power exchange is a new requirement of the European Network Code on Demand and Connection. The controller adjusts the reactive power of the distributed generators and the voltage reference of HV/MV on load tap changers and capacitor banks. This method was simulated on a 20 kV network taking into account actual technical limitations of distribution networks
