Institute for Systems Analysis

Control theory has made significant progress over the years, yet the challenge of selecting an accurate model for controller design remains a critical issue. Even small mismatches between the assumed model and true system dynamics can lead to performance degradation or instability, particularly in Linear Quadratic Gaussian (LQG) control, which relies on precise noise and system modeling. While data-driven approaches have been explored to mitigate model inaccuracies, conventional LQG methods still struggle in non-Gaussian noise environments. To address this limitation, we propose a novel hybrid framework that integrates polynomial-based methods with a data-based Markov parameter approach. By leveraging higher-order statistical information, our method improves the control performance under non-Gaussian noise, extending the applicability of LQG control in real-world scenarios. Simulations highlight the potential of the proposed approach.

This work proposes the Named Entities Semantic Relatedness (NESR) method to compute the semantic relatedness between named entities in DBpedia. Semantic relatedness is applied in many contexts, such as entity linking and word sense disambiguation, and DBpedia is one of the largest publicly available knowledge graphs. Furthermore, named entities, which are resources that represent real-world objects, in general, in DBpedia, have richer descriptions than conceptual entities. Given two named entities, the method computes their degree of semantic relatedness considering weighted reciprocal references of adjacent resources in the paths connecting them. NESR was experimented with five datasets and compared with ten methods to compute semantic relatedness in an RDF knowledge graph. The results show that, concerning Spearman’s correlation, on average, NESR outperforms each method by at least $8 \%$, and outperforms the method that performs best by $11 \%$ on average.

Prime editing is a promising approach for correcting pathogenic variants, but its efficiency remains variable across genomic contexts. Here, we systematically evaluated 12 modifications of the PEmax system for correcting the CFTR F508del pathogenic variant that caused cystic fibrosis in patient-derived airway basal cells. We chose EXO1 and FEN1 nucleases to improve the original system. While all tested variants showed comparatively low efficiency in this AT-rich genomic region, 4-FEN modification demonstrated significantly improved editing rates (up to 2.13 fold) compared to standard PEmax. Our results highlight two key findings: first, the persistent challenge of AT-rich target sequence correction even with optimized editors, and second, the performance of 4-FEN suggests its potential value for other genomic targets.

Lymphangioleiomyomatosis (LAM) is a rare progressive disease that affects women of reproductive age and is characterized by cystic lung destruction, airflow obstruction, and lymphatic dysfunction. Current diagnostic methods are costly or lack sufficient specificity, highlighting the need for novel non-invasive approaches. Exhaled breath analysis using real-time proton mass spectrometry (PTR-MS) presents a promising strategy for identifying disease-specific volatile organic compounds (VOCs). This cross-sectional study analyzed exhaled breath samples from 51 LAM patients and 51 age- and sex-matched healthy controls. PTR-time-of-flight mass spectrometry (PTR-TOF-MS) was employed to identify VOC signatures associated with LAM. Data preprocessing, feature selection, and statistical analyses were performed using machine learning models, including gradient boosting classifiers (XGBoost), to identify predictive biomarkers of LAM and its complications. We identified several VOCs as potential biomarkers of LAM, including m/z = 90.06 (lactic acid) and m/z = 113.13. VOCs predictive of disease complications included m/z = 49.00 (methanethiol), m/z = 48.04 (O-methylhydroxylamine), and m/z = 129.07, which correlated with pneumothorax, obstructive ventilation disorders, and radiological findings of lung cysts and bronchial narrowing. The classifier incorporating these biomarkers demonstrated high diagnostic accuracy (AUC = 0.922). This study provides the first evidence that exhaled breath VOC profiling can serve as a non-invasive additional tool for diagnosing LAM and predicting its complications. These findings warrant further validation in larger cohorts to refine biomarker specificity and explore their clinical applications in disease monitoring and personalized treatment strategies.

Context Pregnancy scanning allows ewes to be separately managed by litter size class to: (1) meet recommended condition score targets to enhance survival and progeny productivity, (2) prioritise allocation of multiple-bearing ewes to smaller mobs or more sheltered lambing paddocks, and (3) select replacement ewes based on birth type. However, 70% of Australian sheep producers are not scanning for litter size. Aims To establish the profitability of pregnancy scanning in southern Australia, to build a stronger business case to address the key perception by non-adopters that the case is weak. Methods The profitability of pregnancy scanning was modelled, across three regions (long, medium and short growing seasons in southern Australia), three different genotypes (Merino, Merino × terminal and a maternal flock) and three lambing times (autumn, winter and spring) using the Australian Farm Optimisation Model. A simpler gross margin analysis was used to include the summer rainfall regions. Key results Implementing optimal ewe nutritional management, paddock allocation and replacement strategies based on litter size was profitable in all the scenarios of region, genotype, time of lambing and price tested. The increase in profitability in the winter rainfall zone averaged across the scenarios tested was A$5.75/ewe scanned. In the summer rainfall zone, the average increase in profit was A$4.44/ewe scanned. Profitability was sensitive to sheep meat prices, but not sensitive to wool and supplement prices, nor overall reproductive rate. Conclusions There is a strong business case for pregnancy scanning ewes for litter size via: (1) selling of non-pregnant ewes (A$1.75/ewe), (2) better allocation of feed based on litter size (A$2.00/ewe), (3) allocating multiple-bearing ewes to the best lambing paddocks and/or reducing mob size to enhance lamb survival (A$1.00/ewe), and (4) accounting for birth type when selecting breeding replacements (A$1.00/ewe). Implications Pregnancy scanning for litter size is a relatively low-cost, high-profit and vital tool for improving reproductive rate, lamb and ewe survival rates, and lamb growth rates. Our findings extend previous, more limited studies, in terms of their reach across different regions, genotypes and production systems.

Quantitative assessment of microcirculation is of paramount importance in the diagnosis and monitoring of a wide range of diseases. This paper presents an image processing algorithm for imaging photoplethysmography sequences, designed for analyzing microcirculation processes with using video photoplethismogram signals. The algorithm comprises two primary blocks: a comprehensive pre-processing stage, including stabilization, illumination correction, and background component removal; and a time-frequency analysis based on the Fourier transform, followed by the calculation of a Root Mean Square Deviation (RMSD) map. The proposed approach enables the identification of pixels corresponding to dynamic intensity variations and can be utilized for the analysis of temporal changes in images. The developed algorithm can be applied in diagnostic tasks requiring analysis of the temporal dynamics of intensity in image arrays. The algorithm’s efficacy was successfully demonstrated in three applications: diagnosis of dermatographic urticaria, investigation of microcirculation during Photodynamic Therapy (PDT), and mapping the microcirculatory network of nevi.

Robotic mobile platforms are key building blocks for numerous applications and cooperation between robots and humans is a key aspect to enhance productivity and reduce labor cost. To operate safely, robots typically rely on a custom map of the environment that depends on the sensor configuration of the platform. In contrast, blueprints stand as an abstract representation of the environment. The use of both CAD and SLAM maps allows robots to collaborate using the blueprint as a common language, while also easing the control for non-robotics experts. In this work we present an adaptive system to project a 2D pose in the blueprint to the SLAM map and vice-versa. Previous work in the literature aims at morphing a SLAM map to a previously available map. In contrast, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CAD2SLAM</i> does not alter the internal map representation used by the SLAM and localization algorithms running on the robot, preserving its original properties. We believe that our system is beneficial for the control and supervision of multiple heterogeneous robotic platforms that are monitored and controlled through the CAD map. Finally, we present a set of experiments that support our claims as well as open-source implementation.