University Džemal Bijedić of Mostar

With the implementation of Industry 4.0 in manufacturing processes and logistics systems, the manipulation of packages with different dimensions, masses, and geometric characteristics presents a significant challenge for automated palletizing within the production process itself. The aim of this paper is the development and optimization of an adaptive gripper intended for collaborative robots, with a focus on reliable retrieval and palletizing of products with variable characteristics. The proposed solution includes a mechanically adjustable gripping mechanism that enables efficient manipulation of boxes with a mass of up to 10 kg, with minimal need for manual adjustment. Within the research, an analysis of technical requirements, the development of a CAD model, the definition of key functionalities, and the optimization of gripper operating parameters were carried out, including gripping force, manipulation stability, and cycle time. Experimental validation showed that the developed gripper significantly increases the success rate of gripping and reduces the risk of package deformation, while integration with a collaborative robot enables safe and flexible application in industrial environments. The results indicate a high potential of adaptive grippers for improving automated palletizing systems and achieving a higher level of efficiency and reliability in operation.

Modern football is characterized by playing both in defense and attack, which requires greater energy demands from players, and the amount and intensity of movement increases from year to year. The aim of the study was to determine differences in the distance covered and movement intensities of players in relation to their position in the team. The sample was players who played all 90 minutes of the knockout phase of the 2022 World Cup (N=224), according to positions: goalkeepers (n=31), defense (n=101), midfielders (n=61), and attackers (n=31). Data were taken from the official FIFA website (www.fifa.com): distance covered (m), distance covered in zone 1 (speed 0-7 km/h), in zone 2 (7-15 km/h), in zone 3 (15-20 km/h), in zone 4 (20-25 km/h) and in zone 5 (>25 km/h), number of runs in zone 4, number of sprints in zone 5, maximum achieved speed (km/h). Differences between positions were determined by discriminant analysis. Three discriminant functions were isolated that are statistically significant at the 99% level (sig.=0.000) (Can. Cor.=0.934; Can. Cor =0.492 and Can. Cor.=0.356). The highest correlations with the first function, which maximally differentiates positions (Wilks Lambda =0.085; sig.=0.000), have the variables: intensity of movement in zone 2, number of runs in zone 4 and total distance covered. The second function (Wilks Lambda =0.662; sig.=0.000) is determined by the maximum achieved speed and distance covered in zone 3. The third function (Wilks Lambda =0.873; sig.0.000) is determined by intensity of movement in zone 4, number of sprints, distance covered in zone 5 and in zone 1. Excluding the goalkeeper position, it is evident that positions in the team are approaching each other in relation to distance covered, intensity of movement and maximum speed of movement, which supports the thesis that in modern football, polyvalent football players who can be used in multiple positions in the team are increasingly profiled

Internet of Things (IoT) nodes deployed in non-high-precision monitoring environments tend to waste a lot of energy with redundant sensor readings, data transfers as well as poor low level hardware utilization, combined with a lack of understanding of the environment they’re deployed in. This paper presents a framework that through the use of a Digital Twin (DT) enables autonomous node management with predictive maintenance enabling the node with self-healing capabilities, designed to minimize the dependency on cloud computing. Employing an event-driven edge layer, the analytical intelligence is delegated to a DT. The system fuses Kalman Filtering with unsupervised Machine Learning (Isolation Forest) to dynamically adapt duty cycles, estimate internal battery resistance without dedicated hardware and filter out normal usage bursts. Through independent testing the use of aforementioned methods showed significant improvement in continuous node operation and battery lifespan expenditure, projected up to 92.1% in comparison to the baseline micro-controller while maintaining near-zero false anomaly reports.

The Iterated Prisoner's Dilemma (IPD) remains a central model for studying the emergence of cooperation, and modern strategy libraries contain hundreds of distinct algorithms whose behavioral relationships are not well characterized by existing taxonomies based on memory depth or implementation details. We present a behavior-driven approach to organizing this strategy space that combines two ideas. First, our procedure adaptively allocates simulation runs across pairs: it samples each match-up until its outcome distribution stabilizes within a target confidence interval, directing computation toward noisy pairs rather than spending it uniformly. Second, each strategy is represented by its full pairwise outcome distribution — the four probabilities (CC, CD, DC, DD) against every opponent in the library — yielding a behavioral fingerprint that distinguishes strategies with similar cooperation rates but different exploitation patterns. We apply three clustering methods (K-means, Ward, spectral) and report inter-method agreement using the Adjusted Rand Index and Normalized Mutual Information, treating partial agreement as evidence about the structure of the strategy space rather than as a single ground truth. Applied to 224 short runtime strategies from the Axelrod library, the analysis recovers a small number of interpretable behavioral archetypes and exposes the cooperation-exploitation tension as an empirical regularity in the cluster-vs-cluster payoff structure. We discuss limitations, including moderate cluster cohesion and partial method agreement, that suggest alternative analyses may legitimately reach different conclusions at higher cluster counts.

This article examines the concept of militant democracy, a particular form of democratic order that actively defends itself against political actors and ideologies aiming to destroy democracy by using democratic institutions. The study focuses on the complex interrelationship between militant democracy and fundamental human rights and freedoms in the context of safeguarding the democratic order. Since militant democracy inherently involves the possibility of restricting certain rights to prevent antidemocratic actions, this article examines the normative justifiability of such limitations. The theoretical analysis traces the evolution of the concept of militant democracy through the works of key political and legal theorists, with special emphasis on Karl Loewenstein and Hans Kelsen. Alongside a conceptual approach, the authors analyse the application and acceptability of the instruments of militant democracy through the practice of competent judicial bodies. The findings reveal that there is no single position on the acceptability of militant democracy in legal theory whereas certain measures of militant democracy, especially the ban on political parties that advocate undemocratic goals, are recognized and accepted in legal practice as legitimate means for preserving democratic institutions.