Horário: 14h
Palestrante: Leonardo Henrique Moreira (doutorado)
Orientadora: Profa Célia Ghedini Ralha
Title: Multi-agent Planning Literature Review
Abstract: Multi-agent planning (MAP) integrates multi-agent systems with practical reasoning agents focusing means-end decision with planning capabilities. MAP computational problem is exponential depending on the number of agents involved in the coordination process, considered NP-hard by some authors. We conducted a literature review with works published from 2014 to 2022 available at Scopus and Web of Science repositories to provide a comprehensive overview of MAP concepts, techniques, and challenges. One focus of the review is the planning process in dynamic environments, where agents can be affected by unexpected, uncontrolled, or non-deterministic events (exogenous), leading to plan failures. In this case, research works generally propose strategies categorized into replanning, repairing, or plan reusing. The consolidated meta-analytic approach (TEMAC) is the literature review methodology, including quantitative techniques and bibliometric aspects. The results highlight the journals and conferences with the highest numbers of related work and the most cited authors. Co-citation and bibliographic coupling clusters categorizing the research efforts were presented, dealing with works cited together or citing the same ones. With the results, we aim to contribute to MAP research by providing a recent overview of concepts, trends and works.
Horário: 14h20
Palestrante: Carlos Joel Tavares da Silva (mestrado)
Orientadora: Profa Célia Ghedini Ralha
Title: An Empirical Architecture Integrating Planning to Self-adaptive Systems: A Multi-Robot Mission Coordination in a Healthcare Service Case
Abstract: Runtime adaptation in self-adaptive systems (SASs) needs planning to mitigate disruptions and continually adjust their behavior in the presence of failures. However, approaches integrating automated planning into SASs are missing establishing barriers to the design of better and tailored autonomous systems. To address this gap, we present a work-in-progress approach introducing an empirical architecture to integrate automated planning to SASs illustrated in the service robot’s domain. Applications of service robots are complex and challenging due to the coordination of heterogeneous agents interacting with humans when performing convenient tasks in real-world environments. We used multi-robot mission coordination in a healthcare service domain case based on the MissionControl ensemble architecture. Our contributions aim to motivate the design and development of automated planning integrated into SASs. The experimental results show that it is possible to generate runtime-adapted plans satisfying the goals of the multi-robot coordination case mitigating mission disruptions.
Horário: 14h40
Palestrante: Aldo Henrique Dias Mendes (doutorado)
Orientadora: Profa Célia Ghedini Ralha
Title: MAS-Cloud+: A Multi-agent Architecture for Optimized Cloud Resource Management
Abstract: Optimized cloud resource management is challenging since resources involve different characteristics, technologies, and financial costs. Multi-agent technologies can offer noticeable improvements for resource management, with intelligent agents deciding on Virtual Machine (VM) resources to perform tasks aiming to reduce time, cost, and waste. This article proposes MAS-Cloud+, an architecture with intelligent agents for predicting, provisioning, and monitoring optimized cloud computing resources. The agents use a hybrid reasoning model that considers three reasoning models: heuristic, formal optimization, and metaheuristic. MAS-Cloud+ instantiates VM considering Service Level Agreement (SLA) on cloud platforms, prioritizing user needs considering time, cost, and waste of resources to perform tasks. To evaluate MAS-Cloud+, we use a DNA sequence comparison application subjected to different workload sizes on the cloud AWS EC2. The choice and provisioning of VMs with the three reasoning models allowed us to evaluate the agents' decisions. In the experiments, our optimization model presented 23.50% better performance and 21.04% less monetary cost than the heuristic one, and 19.27% better performance and 6.53% less cost when compared to the metaheuristic one. Considering the VM random choice, often used by cloud users, MAS-Cloud+ assigned VMs in which the application's executions were up to 74% more cost-effective than setting a single machine with 8 and 96 vCPUs. In addition, a comparative study with state-of-the-art work executed on the AWS EC2 and Google clouds reveals that MAS-Cloud+ decision-making models consistently strive to balance cost, time, and waste. These results indicate that MAS-Cloud+ is a promising solution for optimized cloud resource management.
Horário: 15h
Palestrante: Aurelio Ribeiro Costa (doutorado)
Orientadora: Profa Célia Ghedini Ralha
Title: Adaptive Model to Community Detection in Dynamic Social Networks
Abstract: A vital problem tackled in network analysis is community structure identification. However, the current use of network analysis techniques concentrates on analyzing static community structures, which generates a research gap not considering the dynamic aspects. Some solutions for the community detection problem adapted to the dynamicity of the network present limitations on the resulting performance, and others do not fit such contexts. This situation aggravates when considering the demand to analyze constantly growing social networks. This research aims to fulfill this gap by focusing on the topology change along a time frame, applying deep reinforcement learning to the problem of community detection on dynamic social networks. We propose an adaptive model with an actor-critic reinforcement learning-based architecture to maximize the local modularity density of a community structure using a graph neural network to cope with changing aspects of large social networks. Experiments conducted using the Actor Critic-based architecture to Community Detection (AC2CD) with real-world dynamic social network datasets show accuracy comparable to the state-of-the-art solutions. Although the results indicate that the architecture copes well with real-world social networks, further investigation is necessary to improve computational aspects such as performance, and accuracy considering the unbalancing aspect of the network.
Horário: 15h20
Palestrante: Natan de Souza Rodrigues (doutorado)
Orientador: Profa Célia Ghedini Ralha
Title: Author Name Disambiguation Literature Review with Consolidated Meta-Analytic Approach
Abstract: Name ambiguity is a common problem in many bibliographic repositories affecting data integrity and validity. This article presents an Author Name Disambiguation (AND) literature review using the theory of the consolidated meta-analytic approach, including quantitative techniques and bibliometric aspects. The study covers information from 211 documents from the Web of Science and Scopus databases from 2003 to 2022. We identified that the most widely used approaches for solving AND are probabilistic algorithms, graph-oriented computation, classification, training algorithms, and other recent alternatives. A proposed taxonomy, based on the literature stratification, helped to organize the identified approaches. The countries that publish most in AND area are the USA, China, and Germany, with 21%, 19%, and 13% of the total papers, respectively. The study provides an overview of AND state-of-the-art research and can direct further investigation based on the quantified information from the past.
Local: Teams- Equipe PPGI0095 Seminário, Canal 1-2023
https://teams.microsoft.com/l/channel/19%3a94dc8846a9ca47e389e0083f2b15fe70%40thread.tacv2/Semin%25C3%25A1rios%25201-2023?groupId=93b66213-b249-467a-bcbe-dcd4255edf95&tenantId=ec359ba1-630b-4d2b-b833-c8e6d48f8059
Profa Célia Ghedini Ralha (Este endereço de email está sendo protegido de spambots. Você precisa do JavaScript ativado para vê-lo.)
Coordenadora Seminários de Pós-Graduação em Informática 1-2023