Search

Utilizing parallel algorithms is an established way of increasing performance in systems that are bound to real-time restrictions. Sensor-based sorting is a machine vision application for which firm real-time requirements need to be respected in order to reliably remove potentially harmful entities from a material feed. Recently, employing a predictive tracking approach using multitarget tracking in order to decrease the error in the physical separation in optical sorting has been proposed. For implementations that use hard associations between measurements and tracks, a linear assignment problem has to be solved for each frame recorded by a camera. The auction algorithm can be utilized for this purpose, which also has the advantage of being well suited for parallel architectures. In this paper, an improved implementation of this algorithm for a graphics processing unit (GPU) is presented. The resulting algorithm is implemented in both an OpenCL and a CUDA based environment. By using an optimized data structure, the presented algorithm outperforms recently proposed implementations in terms of speed while retaining the quality of output of the algorithm. Furthermore, memory requirements are significantly decreased, which is important for embedded systems. Experimental results are provided for two different GPUs and six datasets. It is shown that the proposed approach is of particular interest for applications dealing with comparatively large problem sizes.

In the growing field of bulk solids handling, automated optical sorting systems are of increasing importance. However, the initial sorter calibration is still very time consuming and the precise optical sorting of many materials still remains challenging. In order to investigate the impact of different operating parameters on the sorting quality, a numerical model of an existing modular optical belt sorter is presented in this study. The sorter and particle interaction is described with the Discrete Element Method (DEM) while the air nozzles required for deflecting undesired material fractions are modelled with Computation Fluid Dynamics (CFD). The correct representation of the resulting particle–fluid interaction is realized through a one–way coupling of the DEM with CFD. Complex shaped particle clusters are employed to model peppercorns also used in experimental investigations. To test the correct implementation of the utilized models, the particle mass flow within the sorter is compared between experiment and simulation. The particle separation results of the developed numerical model of the optical sorting system are compared with matching experimental investigations. The findings show that the numerical model is able to predict the sorting quality of the optical sorting system with reasonable accuracy.

To reduce the amount of data transfer in networked systems measurements are usually taken only when an event occurs rather than periodically in time. However, a fundamental assessment on the response of estimation algorithms receiving event sampled measurements is not available. This research presents such an analysis when new measurements are sampled at well-designed events and sent to a Luenberger observer. Conditions are then derived under which the estimation error is bounded, followed by an assessment of two event sampling strategies when the estimator encounters two different types of disturbances: an impulse and a step disturbance. The sampling strategies are compared via four performance measures, such as estimation-error and communication resources. The result is a clear insight of the estimation response in an event-based setup.

Hintergrund: Hantaviren gehören zur Familie der Bunyaviridae, in Mitteleuropa ist der Typ Puumala am weitesten verbreitet. Reservoir ist die Rötelmaus, die Infektion des Menschen erfolgt durch Inhalation virushaltiger Aerosole. Hantaviren zeigen eine Affinität zu Lunge und Niere, der [for full text, please go to the a.m. URL], 21. Jahrestagung der Deutschen Gesellschaft für Pädiatrische Infektiologie (DGPI)

To reduce the amount of data transfer in networked systems, measurements are usually taken only when an event occurs rather than periodically in time. However, this complicates estimation problems considerably as it is not guaranteed that new sensor measurements will be sampled. In order to cope with such event sampled measurements, an existing state estimator is modified so that any divergent behavior in estimation results will be curtailed. To start, a general formulation of event sampling is proposed, which is then used to set up a state estimator combining stochastic as well as set-membership measurement information according to a hybrid update: when an event occurs the estimated state is updated using the stochastic measurement received (positive information), while at periodic time instants no new measurement is received (negative information) and the update is based on knowledge that the sensor value lies within a bounded subset of the measurement space. An illustrative example further shows that the developed estimator has an improved representation of estimation errors compared to purely stochastic estimators for various event sampling strategies.

The paper focuses on the fundamental idea of Kalman's seminal paper: how to solve the filtering problem from the only knowledge of the first two moments of the noise terms. In this paper by exploiting set of distributions based filtering we solve this problem without introducing additional assumptions on the distributions of the noise terms (e.g. Gaussianity) or on the final form of the estimator (e.g. linear estimator). Given the moments (e.g. mean and variance) of random variable X it is possible to define the set of all distributions that are compatible with the moments information. This set of distributions can be equivalently characterized by its extreme distributions which is a family of mixtures of Dirac's deltas. The lower and upper expectation of any function g of X are obtained in correspondence of these extremes and can be computed by solving a linear programming problem. The filtering problem can then be solved by running iteratively this linear programming problem.

Colloque avec actes et comité de lecture. Internationale.; International audience; L'assimilation variationnelle de données consiste à combiner les informations issues de simulations numériques à celles provenant de mesures afin de reconstruire de manière optimale l'état du système. En pratique, cette approche est limitée par le coût de la résolution du problème direct/adjoint qui lui est associé. Pour cela, nous évaluerons ici sur une équation de convection/diffusion modèle, différentes stratégies de résolution basées sur des modèles réduits adaptatifs POD. En particulier, nous montrerons l'intérêt de coupler modèles réduits et méthode d'optimisation à région de confiance.

Many modern fusion architectures are designed to process and fuse data in networked systems. Alongside the advantages, such as scalability and robustness, distributed fusion techniques particularly have to tackle the problem of dependencies between locally processed data. In linear estimation problems, uncertain quantities with unknown cross-correlations can be fused by means of the covariance intersection algorithm, which avoids overconfident fusion results. However, for nonlinear system dynamics and sensor models perturbed by arbitrary noise, it is not only a problem to characterize and parameterize dependencies between estimates, but also to find a proper notion of consistency. This paper addresses these issues by transforming the state estimates to a different state space, where the corresponding densities are Gaussian and only linear dependencies between estimates, i.e., correlations, can arise. These pseudo Gaussian densities then allow the notion of covariance consistency to be used in distributed nonlinear state estimation.