http://repositorio.ufc.br/handle/riufc/379 2026-08-14T08:17:49Z http://repositorio.ufc.br/handle/riufc/87415 Título: Treinando agentes de forrageamento multissensoriais usando currículo Autor(es): Férrer Filho, Rômulo Freire Abstract: Deep Reinforcement Learning (DRL) has demonstrated significant success in the development of agents capable of solving complex game tasks. However, the majority of game agents rely solely on visual sensors to gather information about the environment. Recent studies have indicated that agents utilizing audio sensors can outperform those that depend exclusively on vision. This work evaluates the ability to utilize multiple sensors in a foraging task. It also proposes a curriculum-based training strategy to develop agents that effectively use audio as a source of information in foraging scenarios. First, it is demonstrated that agents with both visual and auditory capabilities perform similarly to agents with only visual sensors, indicating that the former ignores the audio input. Initially, it is demonstrated that agents equipped with both vision and hearing capabilities perform similarly to agents with only visual sensors, suggesting that the former tend to ignore the audio input. Subsequently, it is shown that by implementing a gradually increasing difficulty curriculum, the agent effectively utilizes the available audio information, enhancing its robustness in scenarios where visual information is absent. The results indicate that agents can be trained to effectively use audio as a source of information through a curriculum-based training strategy, thereby improving their capacity to handle a wider range of tasks compared to agents with only vision. Tipo: Dissertação 2024-01-01T00:00:00Z http://repositorio.ufc.br/handle/riufc/87189 Título: Modelo de paridade e heurísticas para o problema do caminho positivo mínimo Autor(es): Sombra, João Victor Fonseca Abstract: Given a digraph G = (V,A) with weighted arcs labeled with positive or negative signs, the Minimum Positive Path problem (Shortest Positive Path, SPP) seeks the minimum-cost path between a pair of vertices in G, subject to the constraint that the path contains an even number of negative arcs. Frequently applied in social networks to establish connections between individuals or assess their compatibility, the problem also appears in studies of relationships and influences among biological entities. This work proposes new Integer Linear Programming formulations, improves existing ones through valid inequalities and constraints derived from linear relaxations, and introduces two heuristics for the problem. The new solution approaches were evaluated on synthetic instances of small, medium, and large scale, as well as on real-world instances. For instances with arbitrary real weights, the (SP+) formulation outperforms the models in the literature in terms of running time and enables obtaining optimal solutions on larger-scale digraphs, especially when combined with the bucket heuristic. For instances with strictly positive weights, the new parity model (SPxor), based on an expanded digraph, achieved the best computational performance across all experiments conducted, being on average 94% faster than the models in the literature for the synthetic instances considered. Furthermore, incorporating the proposed heuristics into preexisting formulations consistently reduced solving times, reinforcing their robustness and general applicability. The results obtained allow large-scale SPP instances to be handled with significantly shorter computational times than those reported in the literature. Tipo: Dissertação 2026-01-01T00:00:00Z http://repositorio.ufc.br/handle/riufc/86239 Título: Guia de teste de interoperabilidade para aplicações de Internet das Coisas Autor(es): Branco, Karina da Silva Castelo Abstract: Technology has significantly transformed human interactions with everyday objects, expanding the way they communicate. This broad connectivity with the Internet gave rise to the “Internet of Things (IoT)”, which expanded the boundaries of the Internet to encompass these objects, called smart, which, when interconnected, can collect and share information to provide services effectively. However, the intense data traffic and the diversity of interaction methods of these objects bring several challenges related to interoperability. Interaction methods refer to the different ways in which IoT devices communicate and share data, which can vary widely according to the protocols and standards used. In turn, interoperability refers to the ability of different systems to communicate effectively, ensuring data integrity. In this context, interoperability tests assess the ability of systems and devices to cooperate efficiently. Among the challenges of interoperability testing, the following stand out: architectural complexity, communication between devices, device heterogeneity, and ensuring effective connectivity between them. This master’s dissertation aims to develop an interoperability testing guide for IoT applications based on the methodology proposed by Carvalho et al., (2022). The construction of the guide is based on a literature review, data extraction and analysis, structuring of the guide, and observations of real IoT environments. The guide covers 12 topics, including feature definition and correlation, interoperability testing challenges, environment configuration, subfeatures, contextualization, test cases, measurements, impact of subfeatures, cost-benefit, tool suggestions, and usage examples. The evaluation of the guide consisted of three stages: (1) a structural evaluation using the Technology Acceptance Model (TAM); (2) a controlled experiment applying the guide to test a real IoT application; and (3) an evaluation of the guide with experts. The evaluation showed that the guide provides a comprehensive and practical framework for conducting interoperability testing, assisting experts in identifying problems and improving the integration between IoT systems and devices. Tipo: Dissertação 2024-01-01T00:00:00Z http://repositorio.ufc.br/handle/riufc/85917 Título: On bipolar argumentation frameworks and their equivalence with logic programming and SETAF Autor(es): Cordeiro, Renan de Castro Silva Abstract: Logic programming and abstract argumentation are closely connected paradigms suitable for representing incomplete, contradictory, and/or uncertain information. In Normal Logic Programs (NLPs), each claim is represented by an atom, which is considered true iff it can be derived by a set of logical rules. In Abstract Argumentation Frameworks (AAFs), claims are encoded as arguments and evaluated solely based on an attack relation, which determines how arguments (negatively) interact with each other. Different interactions give rise to distinct argumentation variants, such as Bipolar Argumentation Frameworks (BAFs) and Frameworks with Sets of Attacking Arguments (SETAFs), that additionally allow expressing the notions of support and collective attacks, respectively. For each variant, a evaluation criteria, called semantics, is needed to specify precisely which claims should be regarded as accepted, rejected, or neither (undecided). In this work, we study BAFs by both a semantic and a structural perspective. We propose new semantics for BAFs, called β-semantics, and employ them to find semantic-preserving translations between BAFs, SETAF, and NLPs. Moreover, our proposed translations preserve the structure of these formalisms when restricted to redundancy-free classes of them. The resulting translations are made accessible in the following website: https://web.archive.org/web/20260327154048/https://renpet-hhh.github.io/argnlp/static/. Tipo: Dissertação 2026-01-01T00:00:00Z