Advanced data analytics for integrated port call prediction and optimization

Abstract

We seek enthusiastic candidates that have an interest in and capabilities to perform research on advanced data analytics for integrated port call prediction and optimization models. Although there are several papers that address specific operational problems, such as berth allocation and pilotage planning or container stacking, there is limited research on the integrated modelling of port call operations and controlling total port emissions with data-driven methods. There is extensive work on container terminal operations and design at the seaside, using a variety of OR methods. There are ample opportunities to develop novel data driven methods in this domain. Preliminary research has shown that using publicly available data such as those emitted from automatic identification systems (AIS) can vastly improve the quality of ETA estimates. However, using data-driven optimization methods that make use of such data, when they become readily available to improve call operations is underexplored both in academic research and professional practice.

The PhD student will become an active member of the large and diverse group of researchers at the SCM section. This group is one of the largest of its kind with more than 25 faculty members. It consistently ranks amongst the top three in Europe in terms of research output. Candidates are expected to contribute to the group's world-class research and teaching, and thereby to management science and management practice in logistics and supply chain management. Candidates are also expected to actively engage with peers and port stakeholders in the funded research project and create societal impact.

Keywords

Sea Port, Port Call Optimization, Data Analytics, Operations Research, Machine Learning

Topic

Overall funded research project PortCall.Zero: European seaports must achieve net zero emissions by 2050 and a 55% reduction by 2030 as mandated by the European Green Deal. With up to 80% of port emissions stemming from the port call process, new methods are urgently needed for coordinated decision-making and net-zero strategies. Current methods lack integrated planning, data sharing, and do not address uncertainties from shore power and new fuels. We will develop AI methods to manage these complexities and uncertainties, aiming to decarbonize the port call process. This will be demonstrated in Rotterdam and Moerdijk, making the Netherlands a global leader in sustainable port operations.

PhD project: Although there are several papers that address specific operational problems, such as berth allocation and pilotage planning or container stacking, there is limited research on the integrated modelling of port call operations and controlling total port emissions with data-driven methods. There is extensive work on container terminal operations and design at the seaside, using a variety of OR methods. There are ample opportunities to develop novel data driven methods in this domain. Preliminary research has shown that using publicly available data such as those emitted from automatic identification systems (AIS) can vastly improve the quality of ETA estimates. However, using data-driven optimization methods that make use of such data, when they become readily available to improve call operations is underexplored both in academic research and professional practice.

This project involves: Rob Zuidwijk, Ioannis Fragkos (RSM), Rommert Dekker (ESE), Frederik Schulte (TUD)

Approach

The PhD student will create new anticipatory modelling approaches in two ways. First, research on improving operations will focus on reducing vessel waiting times and improving estimated time of arrival (ETA) through the integration of ML-based models. Second, the focus will go beyond improving existing operations, ultimately aiming to predict, and reduce port call emissions via effective planning. Concretely, leveraging AIS data and data that describe the state of port operations real-time, we aspire to control operations such as the scheduling of vessels, their berth allocation and the moves of secondary equipment using reinforcement learning (RL). In research areas beyond Port Logistics, RL methods have emerged as one of the most promising subfields of machine learning and artificial intelligence in the last decade, and they have managed to achieve superior or human-like performance in a series of tasks.

The PhD candidate is expected to integrate mathematical optimization approaches with deep RL to solve the data-driven stochastic control problems that will emerge. Large-scale optimization methodologies, such as Benders decomposition and column generation have been used successfully to solve complex problems in transportation networks and beyond. At the same time, they cannot handle real-time data and optimize systems whose state evolves over time. Deep RL methodologies have an advantage on this end, but optimizing through RL typically requires excessive computational resources or is infeasible due to large state, action and uncertainty spaces (the curse of dimensionality). In this project, novel approaches from mathematical optimization are expected to be combined with deep RL and lead to effective solutions for the control problems that will emerge from modeling the port operations using real-time data. This may involve, for instance, integration of prediction and optimization (Qi & Shen, 2022) , and a federated use of data analytics methods.

Required profile

We welcome applicants with a quantitative orientation towards problem-solving, such as mathematical modelling (deterministic or stochastic), advanced statistical data analysis, or machine learning. Examples of educational backgrounds include Operations Research, Econometrics, Business Analytics, Applied Mathematics, or Computer Science. This position requires outstanding coding skills and some understanding of stochastic control theory and mathematical optimization. Applicants also have a keen interest interacting with port stakeholders to motivate and inspire their research and help create societal impact.

Required by ERIM

All application documents required by ERIM can be found here.

Expected output

The PhD project should result in a PhD thesis that meets the requirements of the university. In the overarching research project, the following specific outcomes of the PhD student are expected: (1) Establishment of operational and environmental key performance indicators and testing protocols; (2) Emission and performance analysis of existing operations; and (3) Research and development of AI-based decision support methods and systems.

Cooperation

The PhD project will be a cooperative effort with peers and several port stakeholders, among which are colleagues from TU Delft, SmartPort, and several partners in the port community.  PhD candidates are actively encouraged to undertake a research visit to one of the universities in our group’s network, such as MIT, INSEAD, University of Bologna, Georgia Tech, Northwestern University, and HEC Montreal.

Societal relevance

The overarching project PortCall.Zero supports ports to reach net zero emission goals as a collective in 2050. The project provides tools and insights to port managing bodies and industry to assess emission reduction strategies for the port as a whole and coordinate and control the pathway towards net zero port calls. Collaboration is critical to reach net zero targets. The project provides companies with a means to coordinate and share the net effect of all the operations in the port of all the companies, providing a collective benefit. PortCall.Zero turns the energy transition into a big opportunity for the port and the businesses on its premises. By building infrastructures in collaboration, costs and risks are lower. The competitive position of the port is maintained, with companies staying relevant throughout the transition.

Scientific relevance

The proposed research advances knowledge possibly in various directions, including but not limited to: Create new approaches to combine prediction models and decision support modeling and obtain (collect and fuse) reliable data of emissions in the port as a basis of modelling of the port operations’ impact. A challenge is that present AI techniques, that enter in many analytical domains, all require good data sources to train algorithms or to support automated learning or modelling processes. This research will contribute insights into the relationship between data quality and model performance, creating new approaches to combine prediction models and decision support modeling (see, e.g., Bertsimas and Kallus, 2020 or Elmachtoub and Grigas 2022). Since operations are greatly changed with the implementation of new equipment that uses renewable energy (Iris and Lam, 2019), novel modeling approaches need to be found that consider new forms of uncertainties from network effects (e.g., in electrification) or supply uncertainty (e.g., of renewable energy).

Literature references & data sources

Next to the extant body of academic literature, results from our earlier projects with the port feed into the project. We are and have been involved in various national and EU research projects with port stakeholders, such as MAGPIE, Boxreload (TEN-T), CO2REOPT (ERANET), 4TEURN (Kansen voor West), SELIS, PLANET and MAGPIE (Horizon 2020), ISOLA and Trans-SONIC (Topsector Logistics, NWO), but also SURF STAD (Verdus, NWO) and CATALYST (Topsector Logistics, NWO).

Bertsimas, D., & Kallus, N. (2020). From predictive to prescriptive analytics. Management Science, 66(3), 1025-1044.

Elmachtoub, A. N., & Grigas, P. (2022). Smart “predict, then optimize”. Management Science, 68(1), 9-26.

Iris, C. & Lam, J.S.L. (2019). A review of energy efficiency in ports: Operational strategies, technologies and energy management systems. Renewable and Sustainable Energy Reviews, 112(C), 170-182.

Qi, M. & Shen, Z.-J. (2022) Integrating Prediction/Estimation and Optimization with Applications in Operations Management. INFORMS TutORials in Operations Research, published online. https://doi.org/10.1287/educ.2022.0249

Wu, L., Adulyasak, Y., Cordeau, J. F., & Wang, S. (2022). Vessel service planning in seaports. Operations research, 70(4), 2032-2053.

Employment conditions

ERIM offers fully-funded and salaried PhD positions, which means that accepted PhD candidates become employees (promovendi) of Erasmus University Rotterdam. Salary and benefits are in accordance with the Collective Labour Agreement for Dutch Universities (CAO).

Erasmus University Rotterdam aspires to be an equitable and inclusive community. We nurture an open culture, where everyone is supported to fulfil their full potential. We see inclusivity of talent as the basis of our successes, and the diversity of perspectives and people as a highly valued outcome. EUR provides equal opportunities to all employees and applicants regardless of gender identity or expression, sexual orientation, religion, ethnicity, age, neurodiversity, functional impairment, citizenship, or any other aspect which makes them unique. We look forward to welcoming you to our community.

Contact information

For questions regarding the PhD application and selection procedure, please check the Admissions or send us an e-mail via phdadmissions@erim.eur.nl. For enquiries regarding the content of the PhD project, please contact prof. dr. Rob Zuidwijk via rzuidwijk@rsm.nl.