123903 research outputs found
Sort by
Robot planning via LLM proposals and symbolic verification
No full textPlanning in robotics represents an ongoing research challenge, as it requires the integration of sensing, reasoning, and execution. Although large language models (LLMs) provide a high degree of flexibility in planning, they often introduce hallucinated goals and actions and consequently lack the formal reliability of deterministic methods. In this paper, we address this limitation by proposing a hybrid Sense–Plan–Code–Act (SPCA) framework that combines perception, LLM-based reasoning, and symbolic planning. Within the proposed approach, sensory information is first transformed into a symbolic description of the world in Planning Domain Definition Language (PDDL) using an LLM. A heuristic planner is then used to generate a valid plan, which is subsequently converted to code by a second LLM. The generated code is first validated syntactically through compilation and then semantically in simulation. When errors are detected, local corrections can be applied and the process is repeated as necessary. The proposed method is evaluated in the OpenAI Gym MiniGrid reinforcement learning environment and in a Gazebo simulation on a UR5 robotic arm using a curriculum of tasks with increasing complexity. The system successfully completes approximately 71–75% of tasks across environments with a relatively low number of simulation iterations
Effectiveness of office hysteroscopy for retained products of conception
No full textPurpose Retained products of conception (RPOC) are a common complication following pregnancy. Office hysteroscopy (OH) is increasingly used for diagnostics and management due to its minimally invasive nature. However, incidence of incomplete OH removal and procedures in which no RPOC are identified despite prior suspicion remains a concern. This study aimed to identify factors associated with these outcomes to improve patient selection and procedural success.
Methods A retrospective study was conducted on patients referred for OH for presumed RPOC between August 2015 and April 2023 at the Department of Gynecology and Obstetrics, University Medical Centre Ljubljana, Slovenia. Inclusion criteria included hemodynamically stable patients with prolonged bleeding and/or suspicious ultrasound (US) findings post-pregnancy. Patients with RPOC thickness of more than 30 mm or strong tissue vascularization on US were excluded. Data on patient demographics, US features, and procedural outcomes were analyzed.
Results Out of the 468 patients, RPOC removal was performed in 333 cases (71.2%), of which 225 (67.6%) were successfully completed, while in 135 cases (28.8%), the procedure was only diagnostic due to the absence of RPOC. Regarding procedural success, neither the pregnancy outcome, i.e., termination of pregnancy (TOP) vs delivery, nor gestational age significantly correlated with it. Longer time from pregnancy end to OH significantly improved procedural success in both groups, after TOP (p = 0.025) and in cases of large RPOC after delivery (p < 0.001). Parity significantly altered procedural success only in the delivery group (p < 0.007). The success rate of the procedure was significantly higher in cases when only small RPOC were observed (p < 0.001). Absence of RPOC occurred more frequently following TOP than after delivery (p < 0.001). Procedures in which no RPOC were confirmed were significantly associated with a longer interval between pregnancy end and OH in both TOP (p = 0.013) and delivery group (p = 0.003). Gestational age significantly correlated with the absence of RPOC only in the delivery group (p = 0.003). The likelihood of not confirming RPOC was higher where US thickness and length were lower (p = 0.007 and p = 0.011, respectively).
Conclusion OH is effective for managing RPOC with a high success rate, but the absence of RPOC in a considerable number of OH-treated patients stresses the need for better diagnostic criteria and patient selection to minimize overtreatments. Further prospective studies are needed to optimize the timing and indications for OH
Artificial Intelligence and Constitutional Principles
No full textUmetna inteligenca prodira v vsa področja našega življenja. Nedvomno se je izkazala za koristno orodje, a je vendar prinesla s seboj tudi nemalo problematike. Umetna inteligenca izziva ustavna načela Ustave Republike Slovenije, saj jih lahko okrepi ali tudi ogrozi. Načelo demokratičnosti lahko okrepi, saj lahko izboljša delovanje javnih institucij in spodbuja pluralizem, po drugi strani pa ga lahko ogrozi, saj ogroža demokratične volitve in omogoča nadzor nad družbo. Umetna inteligenca se uporablja v vseh treh vejah oblasti, kar postavi pod vprašaj načelo delitve oblasti. Učinkovit sodni sistem je temelj načela pravne države. Uporaba umetne inteligence pri sprejemanju sodnih odločitev lahko ogrozi temeljna jamstva, ki jih zagotavlja Ustava. Glavne težave so neodvisnost, nepristranskost, transparentnost in tudi odgovornost umetne inteligence. Kljub temu je uporaba lahko tudi ključna za varstvo nekaterih drugih jamstev, kot sta sojenje v razumnem času in dostop do sodišča. Nedvomno pa umetna inteligenca vpliva na človekove pravice. Med temi so pravice iz ustvarjalnosti in enakost pred zakonom. Načelo sorazmernosti bo imelo ključno vlogo pri ocenjevanju teh vplivov oz. posegov v človekove pravice. Zaradi vseh teh vplivov je zelo pomembno, da je to področje regulirano. Čeprav v Sloveniji ni sprejete nobene zakonodaje na temo umetne inteligence, je Evropska Unija sprejela prvi celovit pravni okvir o umetni inteligenci na svetu. S tem je bil narejen korak v pravo smer, saj bomo le tako lahko izkoristili prednosti, ki jih ta nova tehnologija prinaša, in se obvarovali pred slabostmi tehnologije.Artificial intelligence is expanding into all areas of our lives. It has undoubtedly proven to be a useful tool but has also brought many issues. Artificial intelligence challenges the constitutional principles of the Constitution of the Republic of Slovenia, as it can strengthen or endanger them. The principle of democracy can be strengthened because artificial intelligence can improve public institutions and support pluralism, but it can also threaten democratic elections and enable control of the society. Artificial intelligence is used in all three branches of government, which questions the principle of separation of powers. An effective judicial system is the foundation of the rule of law. Using artificial intelligence in judicial decisions can endanger the guarantees provided by the Constitution. Key issues include independence, impartiality, transparency, and accountability. However, artificial intelligence can also help protect other guarantees, such as trial within a reasonable time and access to a court. Artificial intelligence clearly affects human rights. These include rights related to copyright law and equality. The principle of proportionality will be crucial in assessing these impacts or interferences with human rights. Because of these effects, regulation is essential. Although Slovenia has not adopted legislation on artificial intelligence, the European Union has passed the world’s first comprehensive legal framework on artificial intelligence. This is a step in the right direction, as it lets us use the advantages of this technology while protecting ourselves from its drawbacks
Development of functional prototype of climbing equipment using 3D printing
No full textNamen diplomskega dela je bil raziskati potencial tehnologije 3D tiska za izdelavo prototipov kompleksne plezalne opreme, specifično plezalne vponke. Osredotočenost je bila na procesu od digitalnega modeliranja do fizične izdelave plastičnega in kovinskega prototipa.
V programskem okolju Fusion 360 je bil najprej zasnovan 3D model vponke, pri čemer so bila upoštevana nekatera načela oblikovanja za dodajalno izdelavo. Nato je bil izbrani model fizično realiziran z uporabo tehnologij modeliranja s spajanjem slojev FDM (angl. Fused Deposition Modelling) in selektivnega laserskega taljenja SLM (angl. Selective Laser Melting). Po končanem tiskanju so bili izvedeni osnovni koraki naknadne obdelave, kot sta odstranjevanje podpornih struktur in površinska obdelava. Izdelani prototipi so bili ovrednoteni z vizualno oceno kakovosti izdelave in kakovosti tiska površine ter z izvedbo dimenzijskih meritev za primerjavo z originalnim CAD modelom.
Z rezultati je bilo potrjeno, da je s tehnologijami 3D tiska, kot sta FDM in SLM, mogoče uspešno izdelati geometrijsko natančen prototip plezalne vponke, ki je uporabna kot fizični model za oceno oblike in prileganja. Čeprav v SLM tehnologiji obstajajo obetavne možnosti za izdelavo kompleksnih kovinskih komponent plezalne opreme, sta za potrditev dejanske funkcionalnosti in varnosti za uporabo v praksi nujni nadaljnja optimizacija procesa in predvsem rigorozno mehansko testiranje v skladu z veljavnimi standardi.This diploma thesis aimed to explore the potential of 3D printing technology for prototyping complex climbing equipment, specifically a climbing carabiner. The focus was on the process from digital modelling to the physical fabrication of both a plastic and a metal prototype.
A 3D model of the carabiner was designed in Fusion 360, incorporating principles of Design for Additive Manufacturing. The model was then physically realised using Fused Deposition Modelling (FDM) and Selective Laser Melting (SLM) technologies. After printing, basic post-processing steps, such as support structure removal and surface finishing, were performed. The manufactured prototypes were evaluated through visual inspection of the build quality and surface finish, as well as by dimensional measurements for comparison with the original CAD model.
The results confirm that 3D printing technologies, such as FDM and SLM, can be used to produce a geometrically accurate prototype of a climbing carabiner, serving as a physical model for form and fit assessment. Although SLM technology offers promising possibilities for manufacturing complex metal components for climbing equipment, further process optimization and, crucially, rigorous mechanical testing in accordance with applicable standards (e.g., UIAA, EN) are essential to confirm actual functionality and safety for practical use
Use of artificial intelligence in Adobe computer programs
No full textNamen diplomskega dela je bil raziskati zmožnosti in omejitve generativne umetne inteligence (UI) pri ustvarjanju vektorskih grafik ter ugotoviti, kako uporabniki prepoznavajo in vrednotijo grafike, ustvarjene s pomočjo UI v primerjavi z deli grafičnih oblikovalcev. Posebna pozornost je bila namenjena vprašanju, ali lahko UI že danes učinkovito nadomesti delo grafičnih oblikovalcev ali del oblikovalskega procesa.
V eksperimentalnem delu smo uporabili funkcijo za pretvarjanje besedila v vektorsko grafiko (angl. text to vector graphic) v programu Adobe Illustrator (različica 28.7.1), kjer smo s pomočjo različnih besedilnih pozivov in dodatnih nastavitev (referenčni slog, učinki, barva in barvni ton) generirali vektorske grafike. Kot dodatno raziskovalno metodo smo izvedli anketo na spletni strani 1ka.si, s katero smo želeli preveriti zaznavanje razlik med grafikami, ustvarjenimi z UI, ter tistimi, ki so jih pripravile bodoče grafične oblikovalke Nina Starman Gombač, Laura Vitežnik, Anika Živec in Tara Baumgarten. Pri tem smo uporabili prenosni računalnik Nitro ANV15-41.
Rezultati ankete so pokazali, da povprečni uporabniki pogosto zaznajo razlike med grafikami, ustvarjenimi z UI, in grafikami, izdelanimi s strani oblikovalcev. Najpomembnejši kriteriji, na katere so se anketiranci opirali, so bili oblika grafike, stopnja detajlov in realističnost. Pri eksperimentih z različnimi funkcijami se je izkazalo, da ima UI največ težav pri zapletenih in večplastnih pozivih, kjer so rezultati pogosto odstopali od pričakovanj.
Zaključujemo, da je UI lahko koristen pripomoček pri oblikovanju, zlasti kot orodje za hitro ustvarjanje idejnih predlogov ali kot pomoč pri vizualizaciji. Kljub temu pa človeška ustvarjalnost in znanje grafičnega oblikovalca ostajata ključna za dosego kakovostnih in vsebinsko ustreznih rešitev.The purpose of this thesis was to explore the capabilities and limitations of generative artificial intelligence (AI) in creating vector graphics and to determine how users recognize and evaluate AI-generated graphics compared to works created by graphic designers. Special attention was given to examining whether AI can already effectively replace the work of graphic designers or certain stages of the design process.
In the experimental part, we used the text-to-vector graphic function in Adobe Illustrator (version 28.7.1) to generate vector graphics from various textual prompts and additional settings (reference style, effects, color, and color tone). As an additional research method, we conducted a survey on the website 1ka.si to examine how users perceive differences between AI-generated graphics and those created by future graphic designers Nina Starman Gombač, Laura Vitežnik, Anika Živec, and Tara Baumgarten. For this purpose, we used a Nitro ANV15-41 laptop.
The survey results showed that average users often notice differences between AI-generated graphics and those produced by designers. The most important criteria participants relied on were the graphic\u27s shape, level of detail, and realism. Experiments with different functions revealed that AI encounters the greatest difficulty with complex, multi-layered prompts, where the results often diverge from expectations.
We conclude that AI can be a useful tool in the design process, especially for quickly generating conceptual proposals or assisting with visualization. Nevertheless, human creativity and graphic designers\u27 expertise remain essential for achieving high-quality, contextually appropriate outcomes
Regional development report as an instrument for measuring regional development and evaluating the achievement of regional policy objectives
No full textOn the Wiener-like root-indices of graphs
No full textIn this paper, we examine roots of graph polynomials where those roots can be considered as structural graph measures. More precisely, we prove analytical results for the roots of certain modified graph polynomials and also discuss numerical results. As polynomials, we use, e.g., the Hosoya, the Schultz, and the Gutman polynomial which belong to an interesting family of degree-distance-based graph polynomialsthey constitute so-called counting polynomials with non-negative integers as coefficients and the roots of their modified versions have been used to characterize the topology of graphs. Our results can be applied for the quantitative characterization of graphs. Besides analytical results on bounds and convergence, we also investigate other properties of those measures such as their degeneracy which is an undesired aspect of graph measures. It turns out that the measures representing roots of graph polynomials possess high discrimination power on exhaustively generated trees, which outperforms standard versions of these indices. Furthermore, a new measure is introduced that allows us to compare different topological indices in terms of structure sensitivity and abruptness
The -irregularity of trees with maximum degree
No full textThe -irregularity, a variant of the well-established Albertson irregularity, is a topological invariant defined for a graph as , where and denote the degrees of vertices and , respectively. Recent research has successfully characterized chemical trees with the maximum -irregularity. In this paper, we expand upon this research by establishing several structural properties of maximal trees with prescribed maximum degree . Application of these properties enables us to characterize maximal trees with . We establish that extremal trees contain only vertices of degrees , and . Moreover, the number of edges with both end-vertices having the degree or is very small, so almost all edges have the (second) maximum possible contribution to -irregularity. We believe this property or similar should extend to maximal trees for any value of , so this is an interesting direction for further research
Photoinduced ring-opening polymerization of N-carboxyanhydrides for the preparation of cross-linked polypeptide gels
No full textUsing light to activate chemical reactions is a powerful approach that allows spatiotemporal control over physicochemical transformations, commonly used in the polymerization and solidification of reactive liquids. Here, we employ a photochemical approach for ring-opening polymerization (ROP) of α-amino acid N-carboxyanhydride (NCA) monomers to prepare covalently cross-linked synthetic polypeptide gels using light. We apply 2-nitrobenzyl carbamate as a photocaged amine (base), where the switch in its activity after illumination is a key factor in controlling ROP. The released dibutylamine, as a basic species, can deprotonate the NCA and thereby initiate polymerization. We investigate the effects of 2-nitrobenzyl dibutyl carbamate on γ-benzyl-L-glutamate NCA polymerization and apply this approach to prepare covalently cross-linked polypeptide gels. We demonstrate excellent 2D spatiotemporal control over gelation, further enhanced by the use of an acid inhibitor that prevents overcuring caused by diffusion of active species from the illuminated region. The development of photoinduced polymerization and cross-linking of NCAs offers significant advantages in polypeptide synthesis and holds great potential for further applications in additive manufacturing
Late Antique and Early Medieval fire strikers in Slovenia
No full textV diplomskem delu sem z arheološkega vidika obravnaval kresila iz poznoantičnih in
zgodnjesrednjeveških kontekstov. Delo temelji na zbiranju obstoječe slovenske literature, ki
sem jo primerjal in dopolnjeval s pomočjo tujih virov. Razpršeno znanje o kresilih sem si
prizadeval združiti v enem celovitem delu, vsa objavljena kresila pa sem zbral v katalogu, kar
omogoča enostavnejše primerjave in analogije. Kresila sem razvrstil v tipe po lastnih merilih
in sestavil okvirno tipokronologijo, obenem pa razrešil in poenotil poimenovanje njihovih
sestavnih delov. Diplomsko delo prispeva k boljšemu razumevanju poznoantičnih in
zgodnjesrednjeveških kresil kot grobnih pridatkov, njihove uporabe in pomena, saj je bilo tem
predmetom v preteklosti posvečene malo pozornosti.In this thesis, I will examine fire strikers from Late Antique and Early Medieval contexts from
an archaeological perspective. The research will be based on collecting existing Slovenian
literature, which I will compare and supplement with foreign sources. I aim to consolidate the
scattered knowledge on fire strikers into a single comprehensive work and compile all published
examples into a catalog, allowing for easier comparisons and analogies. I will classify the fire
strikers into types according to my own criteria and create a preliminary typochronology, while
also clarifying and standardizing the terminology for their individual components. This thesis
contributes to a better understanding of Late Antique and Early Medieval fire strikers as grave
goods, their use, and their significance, as these objects have received little attention in previous
research