Modelamiento estocástico y solución para el scheduling de proyectos de investigación en bases antárticas chilenas.
No Thumbnail Available
Date
2026
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de Concepción
Abstract
Esta tesis desarrolla un marco integrado de modelamiento y resolución para apoyar la toma de decisiones en la planificación de expediciones científicas en la Antártica chilena, abordando de forma conjunta la selección y la programación de proyectos bajo restricciones logísticas extremas, recursos renovables distribuidos por base, relaciones de precedencia y tiempos de traslado interbase. El trabajo se estructura en dos ejes: (i) el Problema de Selección y Programación de Proyecots de Investigación en Múltiples Bases Antárticas (RPSAP, por sus siglas en ingles) determinístico, formulado como un modelo de programación lineal entera mixta que maximiza el beneficio neto (ingresos menos costos de recursos), y (ii) su extensión estocástica de dos etapas, RPSAP estocástico (SRPSAP, por sus siglas en inglés), donde la primera etapa decide la selección y la segunda calendariza por escenario frente a incertidumbre en la duración de proyectos y con incorporación de ventanas temporales. Para el RPSAP se proponen y evalúan metaheurísticas iterated local search (ILS), variable neigborhood search (VNS) y simulated annealing (SA) en 480 instancias, observándose un patrón consistente: ILS tiende a liderar en calidad y SA en rapidez, con ventajas claras en escalas medianas y grandes respecto de enfoques exactos. Para el SRPSAP se diseña un método de descomposición integer L-shaped (IL-S) con estructura maestro–subproblemas, un esquema híbrido IL-S+SA que aproxima las soluciones de segunda etapa mediante SA para acelerar las iteraciones, y una alternativa SA para el problema estocástico (SA-SP) basada en list-scheduling como solucionador directo o mecanismo de inicialización. La validación empírica incluye la adaptación de 36 instancias para el caso estocástico, lo que permite comparar enfoques exactos, metaheurísticos y de descomposición bajo condiciones controladas. En síntesis, los modelos exactos resultan adecuados como referencia y para instancias pequeñas, mientras que la obtención de planes oportunos en escalas realistas requiere estrategias metaheurísticas e híbridas.
This thesis develops an integrated modelling and resolution framework to support decisionmaking in the planning of scientific expeditions in Chilean Antarctica, jointly addressing project selection and scheduling under extreme logistical constraints, renewable resources distributed across stations, precedence relationships, and inter-stations transfer times. The work is structured around two axes: (i) the deterministic Research Project Selection and Scheduling in Multiple Antarctic Station Problem (RPSAP), formulated as a mixed integer programming model that maximises net benefit (revenue minus resource costs), and (ii) its two-stage stochastic extension Stochastic RPSAP (SRPSAP), where the first stage decides the selection and the second schedules by scenario in the face of uncertainty in project duration and with the incorporation of time windows. For the RPSAP, iterated local search (ILS), variable neighborhood search (VNS), and simulated annealing (SA) metaheuristics are proposed and evaluated in 480 instances, with a consistent pattern observed: ILS tends to lead in quality and SA in speed, with clear advantages in medium and large scales over exact approaches. For the SRPSAP, an integer L-shaped decomposition (IL-S) method with a master-subproblem structure is designed, a hybrid IL-S+SA scheme that approximates the second stage solutions using SA to accelerate iterations, and a SA for the stochastic problem (SASP) alternative based on list-scheduling as a direct solver or initialisation mechanism. Empirical validation includes adapting 36 instances to the stochastic case, enabling comparisons of exact, metaheuristic, and decomposition approaches under controlled conditions. In summary, exact models are suitable as a reference and for small instances, while obtaining timely plans at realistic scales requires metaheuristic and hybrid strategies.
This thesis develops an integrated modelling and resolution framework to support decisionmaking in the planning of scientific expeditions in Chilean Antarctica, jointly addressing project selection and scheduling under extreme logistical constraints, renewable resources distributed across stations, precedence relationships, and inter-stations transfer times. The work is structured around two axes: (i) the deterministic Research Project Selection and Scheduling in Multiple Antarctic Station Problem (RPSAP), formulated as a mixed integer programming model that maximises net benefit (revenue minus resource costs), and (ii) its two-stage stochastic extension Stochastic RPSAP (SRPSAP), where the first stage decides the selection and the second schedules by scenario in the face of uncertainty in project duration and with the incorporation of time windows. For the RPSAP, iterated local search (ILS), variable neighborhood search (VNS), and simulated annealing (SA) metaheuristics are proposed and evaluated in 480 instances, with a consistent pattern observed: ILS tends to lead in quality and SA in speed, with clear advantages in medium and large scales over exact approaches. For the SRPSAP, an integer L-shaped decomposition (IL-S) method with a master-subproblem structure is designed, a hybrid IL-S+SA scheme that approximates the second stage solutions using SA to accelerate iterations, and a SA for the stochastic problem (SASP) alternative based on list-scheduling as a direct solver or initialisation mechanism. Empirical validation includes adapting 36 instances to the stochastic case, enabling comparisons of exact, metaheuristic, and decomposition approaches under controlled conditions. In summary, exact models are suitable as a reference and for small instances, while obtaining timely plans at realistic scales requires metaheuristic and hybrid strategies.
Description
Tesis presentada para optar al grado de Doctor/a en Ingeniería Industrial.
Keywords
Planificación, Investigación científica, Scheduling