1. Enhanced precision in axle configuration inference for bridge weigh-in-motion systems using computer vision and deep learningDomen Šoberl, Jan Kalin, Andrej Anžlin, Maja Kreslin, Klen Čopič Pucihar, Matjaž Kljun, Doron Hekič, Aleš Žnidarič, 2025, original scientific article Abstract: Heavy goods vehicles (HGVs) have a significant impact on road and bridge infrastructure, with overloaded vehicles accelerating structural deterioration and increasing safety risks. Bridge weigh-in-motion (B-WIM) systems estimate gross vehicle weight (GVW) using strain measurements, but inaccuracies in axle configuration recognition can reduce reliability. This study presents a low-cost computer vision (CV) extension for existing B-WIM installations that verifies strain-inferred axle configurations using traffic camera images and flags GVW estimates as reliable or unreliable. Experiments on a data set of over 30,000 HGV records show that by combining convolutional neural networks with strain-based heuristics, GVW reliability can improve from 96.7% to 99.89%, effectively excluding nearly all erroneous measurements. The approach operates without interrupting ongoing B-WIM operations and can be applied retrospectively to historical data. Limitations include the inability to detect raised axles (RAs), which the method excludes as unreliable. This method provides a practical, high-precision enhancement for structural health monitoring of bridges.
Keywords: B-WIM, computer vision, deep learning
Published in RUP: 16.01.2026; Views: 204; Downloads: 4
 Full text (2,01 MB)
This document has more files! More... |
2. |
3. |
4. |
5. |
6. Qualitative control learning can be much faster than reinforcement learningDomen Šoberl, Ivan Bratko, 2025, original scientific article Abstract: Reinforcement learning has emerged as a prominent method for controlling dynamic systems in the absence of a precise mathematical model. However, its reliance on extensive interactions with the environment often leads to prolonged training periods. In this paper, we propose an alternative approach to learning control policies that focuses on learning qualitative models and uses symbolic planning to derive a qualitative plan for the control task, which is executed by an adaptive reactive controller. We conduct experiments utilizing our approach on the cart-pole problem, a standard benchmark in dynamic system control. We additionally extend this problem domain to include uneven terrains, such as driving over craters or hills, to assess the robustness of learned controllers. Our results indicate that qualitative learning offers significant advantages over reinforcement learning in terms of sample efficiency, transferability, and interpretability. We demonstrate that our proposed approach is at least two orders of magnitude more sample efficient in the cart-pole domain than the usual variants of reinforcement learning.
Keywords: qualitative modeling, qualitative reasoning, qualitative control, transfer learning
Published in RUP: 07.08.2025; Views: 655; Downloads: 11
Full text (1,53 MB)
This document has more files! More... |
7. Identifying risk factors for sarcopenia using machine learning : insights from multimodal dataFelicita Urzi, Domen Šoberl, Ornella Caputo, Marco Vincenzo Narici, 2025, original scientific article Abstract: Purpose This study aims to identify key risk factors for sarcopenia using machine learning models, leveraging anthropomet- ric, biochemical, functional, nutritional, and genetic data. By developing predictive models, the research seeks to improve early detection, enhance diagnostic accuracy, and facilitate personalized interventions for individuals at risk of sarcopenia. Methods We analysed multimodal data from 484 older adults. Two scenarios: Set-a (including SARC–CalF, excluding SARC-F) and Set-b (including SARC-F, excluding SARC–CalF) were applied in a three-stage modeling process with progressively reduced features and optimized predictive performance using machine learning models. Key predictors were ranked using SHAP values, and model performance was evaluated using AUC, accuracy, sensitivity, and specificity. Internal validation and DeLong’s test were applied to assess robustness and statistical differences. Results The most predictive risk factors included functional measures (chair stand, gait speed), nutritional indicators (pro- tein, folate, copper, vitamin B7), clinical conditions (diabetes, comorbidities, low-density lipoprotein (LDL)), and anthro- pometric markers (body mass index (BMI), calf circumference). Genetic features also contributed to risk stratification. The best-performing model Set-b (with screening test SARC-F) achieved an AUC of 0.951 and an accuracy of 93.62%. While SARC–CalF showed higher individual feature importance, the model achieved an AUC of 0.945 and accuracy of 92.2%. Conclusions This study highlights that traditional sarcopenia screening can be enhanced by capturing complex interplay of functional, nutritional, clinical, and genetic factors, offering clinicians a more accurate and tailored tool for early detec- tion and risk stratification. Future research should focus on validating these models in larger, independent, and longitudinal cohorts to assess their predictive utility across diverse populations and over time.
Keywords: genetics, nutrition, risk factors, sarcopenia
Published in RUP: 23.07.2025; Views: 695; Downloads: 13
Full text (965,36 KB)
This document has more files! More... |
8. |
9. |
10. |
