1. Analysis of the Shortest Path Method Application in Social NetworksBoštjan Šumak, Maja Pušnik, 2023, independent scientific component part or a chapter in a monograph Abstract: This paper analyzes the shortest path problem (SPP) in social networks, based on the investigation and implementation of different methods on a simulated example. The objectives of the paper include identification of the most commonly used methods for finding the shortest path in a social network as a strategic attempt to speed the search of network nodes, focusing on the application of the two most used SPP methods: the Dijkstra and Bellman-Ford algorithms. A comparative analysis is used as an investigation method for performance evaluation of different algorithms, based on their implementation and behavior, tested on a social network example. The research results indicate that the Dijkstra algorithm is faster, and therefore more suitable for searching the shortest connection in social networks.
Keywords: social networks, social networks analysis, shortest path problem, Dijkstra algorithm, Bellman-Ford algorithm, optimization of search
Published in DKUM: 20.01.2026; Views: 0; Downloads: 0
 Full text (2,50 MB)
This document has many files! More... |
2. Multiparametric ▫$Cost–CO_2$▫ optimization of bored reinforced-concrete piles under combined loading in cohesive soilsPrimož Jelušič, 2025, original scientific article Abstract: Laterally loaded slender piles present a classic soil–structure interaction problem where pile displacements and flexural demands are governed by the mobilized lateral resistance of the surrounding soil and the axial-bending capacity of the reinforced concrete section. In response to increasing pressure to reduce embodied emissions, this study develops LAVERCO, an optimization framework for cost- and CO2-efficient design of bored reinforced-concrete piles in cohesive soils subjected to combined lateral and axial actions. The framework integrates Eurocode-based geotechnical checks with full N–M section verification of the RC pile and applies a genetic algorithm over a multi-parametric grid of lateral load, vertical load, and undrained shear strength, using economic cost and embodied CO2 as alternative single objectives. Rank-based (Spearman) sensitivity analysis quantifies how actions, soil strength, and design variables influence the optimal solutions. The results reveal two consistent geometry regimes: CO2-optimal piles are systematically longer and slimmer, while COST-optimal piles are shorter and thicker. In both cases, the objective is dominated by pile length and is reduced by higher undrained shear strength; vertical load has a moderate direct effect, while horizontal load contributes mainly through deflection and bending checks. Feasibility improves significantly in stronger clays, and CO2-optimal geometries generally incur higher costs, clarifying the trade-off between economic and environmental performance. The framework provides explicit geometry-level guidance for selecting bored pile designs that balance cost and embodied CO2 across a wide range of soil and loading conditions and can be directly applied in both preliminary and detailed designs.
Keywords: laterally loaded pile, reinforced-concrete piles, structural analysis, reinforced-concrete design, optimization, CO2 emissions, genetic algorithm, multiparametric analysis, civil engineering practice
Published in DKUM: 19.01.2026; Views: 0; Downloads: 4
Full text (6,87 MB)
This document has many files! More... |
3. Reinforcement learning for robot manipulation tasks in human-robot collaboration using the CQL/SAC algorithmsA. Husaković, Lejla Banjanović-Mehmedović, A. Gurdić-Ribić, Naser Prljača, Isak Karabegović, 2025, original scientific article Abstract: The integration of human-robot collaboration (HRC) into industrial and service environments demands efficient and adaptive robotic systems capable of executing diverse tasks, including pick-and-place operations. This paper investigates the application of Soft Actor-Critic (SAC) and Conservative Q-Learning (CQL)—two deep reinforcement learning (DRL) algorithms—for the learning and optimization of pick-and-place actions within HRC scenarios. By leveraging SAC’s capability to balance exploration and exploitation, the robot autonomously learns to perform pick-and-place tasks while adapting to dynamic environments and human interactions. Moreover, the integration of CQL ensures more stable learning by mitigating Q-value overestimation, which proves particularly advantageous in offline and suboptimal data scenarios. The combined use of CQL and SAC enhances policy robustness, facilitating safer and more efficient decision-making in continually evolving environments. The proposed framework combines simulation-based training with transfer learning techniques, enabling seamless deployment in real-world environments. The critical challenge of trajectory completion is addressed through a meticulously designed reward function that promotes efficiency, precision, and safety. Experimental validation demonstrates a 100 % success rate in simulation and an 80 % success rate on real hardware, confirming the practical viability of the proposed model. This work underscores the pivotal role of DRL in enhancing the functionality of collaborative robotic systems, illustrating its applicability across a range of industrial environments.
Keywords: human-robot collaboration, robot learning, deep reinforcement learning, soft actor-critic algorithm, Conservative Q-learning, robot manipulation tasks
Published in DKUM: 16.01.2026; Views: 0; Downloads: 0
Full text (1,75 MB)
This document has many files! More...
This document is also a collection of 1 document! |
4. Improving mutation strategies in differential evolution with a new pbest selection mechanismJan Popič, Borko Bošković, Janez Brest, 2025, original scientific article Abstract: Differential evolution, which belongs to a group of population-based algorithms, has received a lot of research attention since its introduction in 1995. A population-based algorithm is required to guide individuals to visit potentially better basins of attraction in the search space when searching for a globally optimal solution. Additionally, individuals need to interact with each other during an evolutionary process to explore the search space effectively. In this paper, we propose a novel pbest selection mechanism for DE/current-to-pbest mutation strategy and its variants designed to enhance the potential for exploration of different attraction basins. The proposed mechanism enforces a minimal distance between the selected pbest individual and all other better individuals. This means that possible candidates for the pbest individual, used in mutation, are further spaced apart. As a result, the likelihood that the new trial vector will be generated in a different attraction basin of the search space is increased. The mechanism is incorporated into the L-SHADE, jSO, and L-SRTDE algorithms, and its effectiveness is evaluated using CEC’24 benchmark functions. Experimental results demonstrate improvements in the performance of the selected algorithms, particularly in higher-dimensional problem instances.
Keywords: population-based algorithm, differential evolution, gobal optimization, mutation strategies, exploration–exploitation
Published in DKUM: 29.10.2025; Views: 0; Downloads: 7
Full text (3,09 MB) |
5. Mobile robot localization based on the PSO algorithm with local minima avoiding the fitness functionBožidar Bratina, Dušan Fister, Suzana Uran, Izidor Mlakar, Erik Rot Weiss, Kristijan Korez, Riko Šafarič, 2025, original scientific article Abstract: Localization of a semi-humanoid mobile robot Pepper is proposed based on the particle swarm optimization algorithm (PSO) that is robust to the disturbance perturbations of LIDAR-measured distances from the mobile robot to the walls of the robot real laboratory workspace. The novel PSO, with the avoiding local minima algorithm (PSO-ALM), uses a novel fitness function that can prevent the PSO search from trapping into the local minima and thus prevent the mobile robot from misidentifying the actual location. The fitness function penalizes nonsense solutions by introducing continuous integrity checks of solutions between two different consecutive locations. The proposed methodology enables accurate and real-time global localization of a mobile robot, given the underlying a priori map, with a consistent and predictable time complexity. Numerical simulations and real-world laboratory experiments with different a priori map accuracies have been conducted to prove the proper functioning of the method. The results have been compared with the benchmarks, i.e., the plain vanilla PSO and the built-in robot’s odometrical method, a genetic algorithm with included elitism and adaptive mutation rate (GA), the same GA algorithm with the included ALM algorithm (GA-ALM), the state-of-the-art plain vanilla golden eagle optimization (GEO) algorithm, and the same GEO algorithm with the added ALM algorithm (GEO-ALM). The results showed similar performance with the odometrical method right after recalibration and significantly better performance after some traveled distance. The GA and GEO algorithms with or without the ALM extension gave us similar results according to the accuracy of localization. The optimization algorithms’ performance with added ALM algorithms was much better at not getting caught in the local minimum, while the PSO-ALM algorithm gave us the overall best results
Keywords: mobile robot localization, PSO algorithm, avoid the global minima
Published in DKUM: 17.10.2025; Views: 0; Downloads: 8
Full text (2,95 MB) |
6. Threshold adaptation for improved wrapper-based evolutionary feature selectionUroš Mlakar, Iztok Fister, Iztok Fister, 2025, original scientific article Abstract: Feature selection is essential for enhancing classification accuracy, reducing overfitting, and improving interpretability in high-dimensional datasets. Evolutionary Feature Selection (EFS) methods employ a threshold parameter � to decide feature inclusion, yet the widely used static setting �=0.5 may not yield optimal results. This paper presents the first large-scale, systematic evaluation of threshold adaptation mechanisms in wrapper-based EFS across a diverse number of benchmark datasets. We examine deterministic, adaptive, and self-adaptive threshold parameter control under a unified framework, which can be used in an arbitrary bio-inspired algorithm. Extensive experiments and statistical analyses of classification accuracy, feature subset size, and convergence properties demonstrate that adaptive mechanisms outperform the static threshold parameter control significantly. In particular, they not only provide superior tradeoffs between accuracy and subset size but also surpass the state-of-the-art feature selection methods on multiple benchmarks. Our findings highlight the critical role of threshold adaptation in EFS and establish practical guidelines for its effective application.
Keywords: feature selection, evolutionary algorithm, feature threshold, evolutionary feature selection
Published in DKUM: 14.10.2025; Views: 0; Downloads: 4
Full text (12,73 MB) |
7. Using data mining to improve decision-making : case study of a recommendation system developmentHyrmet Mydyti, Arbana Kadriu, Mirjana Pejić Bach, 2023, original scientific article Abstract: Background and purpose: This study aims to provide a practical perspective on how data mining techniques are used in the home appliance after-sales services. Study investigates on how can a recommendation system help a customer service company that plans to use data mining to improve decision making during its digital transformation process. In addition, study provides a detailed outline on the process for developing and analyzing platforms to improve data analytics for such companies. Methodology: Case study approach is used for evaluating the usability of recommendation systems based on data mining approach in the context of home appliance after-sales services. We selected the latest platforms based on their relevance to the recommender system and their applicability to the functionality of the data mining system as trends in the system design. Results: Evaluation of the impact on decision making shows how the application of data mining techniques in organizations can increase efficiency. Evaluation of the time taken to resolve the complaint, as a key attribute of service quality that affects customer satisfaction, and the positive results achieved by the recommendation system are presented. Conclusion: This paper increases the understanding of the benefits of the data mining approach in the context of recommender systems. The benefits of data mining, an important component of advanced analytics, lead to an increase in business productivity through predictive analytics. For future research, other attributes or factors useful for the recommender systems can be considered to improve the quality of the results. Acknowledgement: The author Hyrmet Mydyti’s PhD thesis has been extended in this paper.
Keywords: digital transformation, data mining, decision tree algorithm, decision-making, home appliances after-sales services
Published in DKUM: 08.10.2025; Views: 0; Downloads: 3
Full text (1,95 MB)
This document has many files! More... |
8. Performance comparison of single-objective evolutionary algorithms implemented in different frameworksMiha Ravber, Marko Šmid, Matej Moravec, Marjan Mernik, Matej Črepinšek, 2025, original scientific article Abstract: Fair comparison with state-of-the-art evolutionary algorithms is crucial, but is obstructed by differences in problems, parameters, and stopping criteria across studies. Metaheuristic frameworks can help, but often lack clarity on algorithm versions, improvements, or deviations. Some also restrict parameter configuration. We analysed source codes and identified inconsistencies between implementations. Performance comparisons across frameworks, even with identical settings, revealed significant differences, sometimes even with the authors’ own code. This questions the validity of comparisons using such frameworks. We provide guidelines to improve open-source metaheuristics, aiming to support more credible and reliable comparative studies.
Keywords: metaheuristics, evolutionary algorithm, metaheuristic optimization framework, algorithm comparison, benchmarking
Published in DKUM: 02.10.2025; Views: 0; Downloads: 4
Full text (3,89 MB) |
9. PrProgramming industrial robots for milling aplications using off-line programming softwares (OLP)Domen Šošter, 2025, undergraduate thesis Abstract: This thesis explores the feasibility of using industrial robots for milling applications, focusing on the challenges of maintaining accuracy. The Yaskawa GP8 robot was used for milling soft materials, and the study emphasizes the importance of offline programming (OLP) for improving precision. The research shows that while robotic milling is viable for soft materials, milling rigid materials would require significant adjustments, including enhanced compensation techniques and calibration. Despite the limited time during the internship, the compensation algorithm was explored and successfully implemented in milling operations. This allowed for measurable improvements in the milling results, demonstrating the potential of the compensation algorithm in enhancing precision.
Keywords: Robotic milling, CAD, CAM, Off-line programming tools, path-correction algorithm
Published in DKUM: 22.09.2025; Views: 0; Downloads: 10
Full text (4,21 MB) |
10. Efficient direct reconstruction of bipartite (multi)graphs from their line graphs through a characterization of their edgesDrago Bokal, Janja Jerebic, 2025, original scientific article Abstract: We study the line graphs of bipartite multigraphs, which naturally arise in combinatorics,
game theory, and applications such as scheduling and motion planning. We introduce
a new characterization of these graphs via valid partial assignments of the edges of the
underlying bipartite multigraph to the vertices of its line graph. We show that an empty
assignment extends to a complete one precisely when the graph is a line graph of a bipartite
multigraph. Based on this, we design an O(∆(G)|E(G)|) algorithm that incrementally
constructs such assignments. The algorithm also provides a data structure supporting
efficient solutions to problems of maximum clique, maximum weighted clique, minimum
clique cover, chromatic number, and independence number. For line graphs of bipartite
simple graphs these problems become solvable in linear time, improving on previously
known polynomial-time results. For general bipartite multigraphs, our method enhances
the O(|V(G)|
3
) recognition algorithm of Peterson and builds on the results of Demaine et al.,
Hedetniemi, Cook et al., and Gurvich and Temkin.
Keywords: UNO-graph, line graph, bipartite graph, bipartite multigraph, graph algorithm
Published in DKUM: 09.09.2025; Views: 0; Downloads: 4
Full text (348,95 KB)
This document has many files! More... |
