Titulo Estágio
Modelling of a real-world berth allocation problem for meta-heuristics
Áreas de especialidade
Sistemas Inteligentes
Engenharia de Software
Local do Estágio
Algorithms and Optimization Laboratory (ALGO), CISUC, DEI-FCTUC
Enquadramento
Berth allocation problems are concerned with the allocation of space, schedules, and terminal resources for vessels in container terminals. Solving these problems efficiently is relevant to companies in the maritime sector to make efficient use of their resources. Since different companies have different needs and data, many different variants of this problem have been studied in the literature.
Berth allocation problems are often solved using mixed-integer linear programming (MILP) models. However, these models may not allow large scale problems to be solved efficiently. In such cases, meta-heuristics offer an exciting alternative. While meta-heuristics for various berth allocation problems have been proposed in the literature, systematic studies comparing them on any number of variants of the problem appears to be lacking. Such a study would be relevant to better understand the performance and trade-offs of different approaches.
Objetivo
The main objective of this work is to model a real-world berth allocation problem relevant to Maersk following a structured modelling approach for meta-heuristics, e.g., using the ROAR-NET API. Another important objective is to study the performance of different meta-heuristics on realistic instances of the problem of interest.
Plano de Trabalhos - Semestre 1
1. Familiarisation with the state of the art in berth allocation, including different modelling approaches
2. Familiarisation with meta-heuristics
3. Development of models for simple versions of the berth allocation problem
4. Intermediate report writing
Plano de Trabalhos - Semestre 2
1. Development of models for more advanced versions of the berth allocation problem
2. Construction of benchmark instances
3. Experimental study
4. Dissertation writing
Condições
Very good background in algorithms and data structures, and at least some familiarity with optimisation and meta-heuristics. A compute server is available to support the experimental work.
Observações
The work will be carried out in collaboration with Maersk in the scope of COST Action Randomised Optimisation Algorithms Research Network (ROAR-NET), CA221137, supported by COST (European Cooperation in Science and Technology).
Co-supervisor: Alexandre D. Jesus
Orientador
Carlos M. Fonseca
cmfonsec@dei.uc.pt 📩