Desenvolvimento de uma DSL para criação de sistemas multiagentes em Python

• Autor(a) principal: Fidler, Igor Guilherme
• Data de Publicação: 2024
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UTFPR (da Universidade Tecnológica Federal do Paraná (RIUT))
• Texto Completo: http://repositorio.utfpr.edu.br/jspui/handle/1/37313
• topic: Linguagem de programação de domínio específico (Computadores); Aprendizado do computador; Framework (Arquivo de computador); Python (Linguagem de programação de computador); Domain-specific programming language; Machine learning; Framework (Computer file); Python (Computer program language); CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

Multi-agent systems consist of a set of agents that cooperate or compete with each other to accomplish a task. This type of system has applications in various areas of knowledge. Several tools have been proposed to facilitate the development of such systems, including the MASPY framework. This framework enables the creation of agents based on beliefs, desires, and intentions and integration with reinforcement learning algorithms in the Python language. Some structures within the framework, such as the plan context and manipulation of the mental state, could be more flexible, especially when compared to agent programming languages. To address this, this work extended the capabilities of MASPY’s plan contexts. It developed a domain-specific language (DSL) that facilitates the manipulation of agents’ mental states and allows the abstraction of the implementation details of a plan context. This work defined the language’s grammar, semantics, and evaluation model and proposed four algorithms for translating the DSL to Python. The results indicate that the DSL shows promise for development with MASPY. However, a limitation found in this work is the need for more clarity about whether the elements proposed in the DSL are sufficient for most uses of the framework; a more in-depth study is needed to verify this. Additionally, there are other open gaps, such as improvements in reported errors and optimization of the generated contexts.