| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 5.04 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Esta dissertação propõe e avalia uma abordagem integrada para o planeamento da pilotagem
marítima que combina engenharia de dados, previsão de tempos e otimização por programação linear
inteira mista (MILP). Trabalhando sobre registos reais de operação portuária — com acesso parcial
a variáveis — o estudo normaliza e valida temporalmente as manobras, enriquece-as com informação
de maré e constrói métricas operacionais consistentes. Para colmatar lacunas informativas, estimase
a duração “a bordo” e da manobra através de modelos preditivos e introduzem-se assunções
explícitas testadas em famílias de cenários.
O modelo de otimização escalona tarefas requeridas de pilotagem respeitando janelas operacionais,
disponibilidade e descanso de pilotos, tempos de deslocação por lancha (derivados de uma topologia
geográfica do porto), e a não-sobreposição de recursos e prioridades. O desenho inclui duas
dimensões distintivas: (i) síntese de turnos on-duty a partir da procura e (ii) termos de função-objetivo
configuráveis para equidade entre pilotos e preferência ambiental.
A avaliação compara soluções do MILP com a realização histórica e com uma regra first-come-firstserved
(FCFS), usando indicadores como espera, atraso face ao horário desejado, pontualidade, uso
de meios externos e equilíbrio de carga de trabalho. Nos cenários testados, a abordagem reduz de
forma consistente espera e atraso, aumenta a pontualidade, diminui dependência de recursos externos
e melhora a equidade, quantificando os compromissos serviço–custo e o efeito de preferências
ambientais.
O contributo é duplo: metodológico, ao articular previsão e otimização com geração de turnos e
tempos de lancha baseados em topologia; e, simultaneamente, prático ao oferecer um pipeline
reexecutável e auditável que transforma dados operacionais em planos de ação. A solução é
transferível para outros portos com adaptações mínimas e abre caminho a integrações em tempo real
e calibração contínua.
This dissertation proposes and assesses an integrated approach to maritime pilotage planning that combines data engineering, time prediction, and mixed-integer linear programming (MILP). Using real port operational records—with partial access to variables—the study performs temporal validation and normalisation of manoeuvres, enriches them with tidal information, and builds consistent operational metrics. To bridge information gaps, on-board and manoeuvre durations are estimated via predictive models, and explicit assumptions are stress-tested across scenario families. The optimisation model schedules inbound/outbound/shift manoeuvres within operational windows, pilot availability and rest rules, pilot-boat travel times (derived from a port topology), resource nonoverlap, and operational priorities. Two distinctive components are included: (i) on-duty shift synthesis from demand when historical rosters are incomplete, and (ii) configurable objective terms for workforce fairness and environmental preference. Evaluate MILP solutions against historical realisations and a first-come-first-served (FCFS) rule using key performance indicators such as waiting time, tardiness relative to desired times, on-time performance, reliance on external resources, and workload balance. Across tested scenarios, the approach consistently reduces waiting and tardiness, improves punctuality, lowers external resource usage, and enhances fairness, while quantifying service–cost trade-offs and the effect of environmental preferences. The contribution is twofold: methodological, by coupling prediction and optimisation with shift synthesis and topology-based launch travel times; and practical, by delivering a reproducible, auditable pipeline that turns operational data into actionable schedules. The solution is transferable to other ports with minor adaptations and paves the way for real-time integration and continuous calibration.
This dissertation proposes and assesses an integrated approach to maritime pilotage planning that combines data engineering, time prediction, and mixed-integer linear programming (MILP). Using real port operational records—with partial access to variables—the study performs temporal validation and normalisation of manoeuvres, enriches them with tidal information, and builds consistent operational metrics. To bridge information gaps, on-board and manoeuvre durations are estimated via predictive models, and explicit assumptions are stress-tested across scenario families. The optimisation model schedules inbound/outbound/shift manoeuvres within operational windows, pilot availability and rest rules, pilot-boat travel times (derived from a port topology), resource nonoverlap, and operational priorities. Two distinctive components are included: (i) on-duty shift synthesis from demand when historical rosters are incomplete, and (ii) configurable objective terms for workforce fairness and environmental preference. Evaluate MILP solutions against historical realisations and a first-come-first-served (FCFS) rule using key performance indicators such as waiting time, tardiness relative to desired times, on-time performance, reliance on external resources, and workload balance. Across tested scenarios, the approach consistently reduces waiting and tardiness, improves punctuality, lowers external resource usage, and enhances fairness, while quantifying service–cost trade-offs and the effect of environmental preferences. The contribution is twofold: methodological, by coupling prediction and optimisation with shift synthesis and topology-based launch travel times; and practical, by delivering a reproducible, auditable pipeline that turns operational data into actionable schedules. The solution is transferable to other ports with minor adaptations and paves the way for real-time integration and continuous calibration.
Descrição
Palavras-chave
Maritime Pilotage Optimization Operational Planning Mixed-Integer Linear Programming (MILP) Resource Scheduling Predictive Time Models Pilotagem Marítima Otimização Planeamento Operacional Programação Inteira Mista (MILP) Escalonamento de Recursos Modelos Preditivos de Tempos
