Asignación de espacios en tiempo real por medio de un modelo basado en agentes (ABM)

Currently, the search for study spaces on a university campus is one of the biggest problems faced by a student because this population, in many cases, exceeds the physical resources of the campuses. Linear models are not feasible to face these types of situations since the large volume of information implied by the study spaces and groups generates data overflow. An alternative to address these situations is a multiagent model which allows assignments dynamically and in real time through interactions between agents, a methodology that due to its characteristics allows progress towards the implementation of a Smart Campus. A model of agents makes possible a “conversation” between agents representing study spaces and groups of students, achieving an assignment similar to what usually happens in real life. It is necessary to highlight that because it is a dynamic model, new groups can be entered even when the model has already been running for a certain time. This implies that, in the long run, the agent model should not be rerun if there is any variation, which is not the case in a linear model. For this reason, this study aimed to implement a multiagent model in the Eclipse development platform for the simulation of different scenarios to assign study spaces to groups of students in real time. To validate the operation of the multiagent model, capacity tests were carried out that sought to know the limitations regarding the amount of data that could be entered, as well as optimality tests in which the objective was to compare the results obtained from an agent model with those of a linear model. Throughout the development of this study, the agent model was compared with a linear model and it was found that this provided results close to the level of a model that could obtain an optimal solution. The results obtained, both in the linear model and in the multiagent model, were subjected to a validator to determine if the results were correct. For a second stage, it was decided to activate the negotiation function, which allowed the groups to give in on their tool requirements in real time. On this occasion, the results were even higher to those of the optimal model, since when input parameters were changed while the model was running, alternative solutions were found that allowed groups to access spaces that were still available, and that in a first instance its assignment was not possible.

• Autores: Cárdenas Garnica, Daniel Alberto; Covo Guzmán, Juliana; Suárez Carrizosa, Natalia; Rodríguez Cruz, Juan Camilo
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2020
• Institución: Pontificia Universidad Javeriana
• Repositorio: Repositorio Universidad Javeriana
• Idioma: spa
• OAI Identifier: oai:repository.javeriana.edu.co:10554/53057
• Acceso en línea: http://hdl.handle.net/10554/53057
• Palabra clave: Asignación de espacios; Multiagentes; Simulación; Smart campus; Agent based model; Simulation; Contract-Net; JADE; Ingeniería industrial - Tesis y disertaciones académicas; Simulación por computadores; Mejoramiento de procesos
• Rights: openAccess
• License: Atribución-NoComercial-SinDerivadas 4.0 Internacional