Conference Papers 2024

Seminar
Day 1 (18 Sep 2024), Session 2, Ports, 14:00 - 15:30

Status
Accepted, awaiting documents

Submitted by / Abstract owner
Mehran Farzadmehr

Authors
Mehran Farzadmehr, Antwerp university (presenter)
Valentin Carlan, Antwerp university
Thierry Vanelslander, Antwerp university

Short abstract
The results of this exercise show that sea carriers, terminals, and nautical chain operators face most challenges that can be addressed by AI. Deep Learning (DL) emerges as the most effective AI method for challenges like ship collision avoidance.

Abstract
The integration of Artificial Intelligence (AI) technologies into the port and shipping industry has started a transformative era. These technologies promise increase of efficiency, optimization, and sustainability. With the world's economy heavily reliant on global trade, ports serve as critical hubs for the movement of goods and commodities (Jeevan et al., 2015). In parallel, the complexities of managing vast logistics operations, having a lack of workforce, facing the impossibility of labor to manage multiple tasks, ensuring the smooth flow of cargo, the need for tools to assist and do double checks in the back-end, and addressing environmental concerns have made this industry see the potential of AI-driven innovations (Filom et al., 2022).
AI technologies are being utilized to address these challenges directly. To name but a few, Machine learning (ML) algorithms are predicting cargo demand for supply chain logistics, computer vision is revolutionizing cargo handling and security by image recognition techniques. These advancements are not only driving down costs and improving the overall operational efficiency of ports and shipping companies, but are also contributing to a more sustainable and environmentally responsible industry (Atak et al., 2021).
This research seeks to illustrate the potential of AI-driven solutions within the port and shipping industry by providing an overview of the challenges that can be effectively addressed through these types of solutions. This overview emphasizes the specific stakeholders and operational areas to which these challenges are relevant. This way, collecting information about AI technologies could help port stakeholders realize where AI can help. The overview also provides valuable information about the requirements of AI technologies in terms of data and computation power for solving challenges. Hence, stakeholders would gain knowledge about, first, whether their challenges can be solved with AI technologies, and second, what type of AI technologies they should consider adopting.
Two individual methods have been employed: a comprehensive literature review of academic publications and interviews with AI developers actively involved in the port and shipping industries. These combined methods gather insights into port-related challenges that can be tackled by availing AI-driven solutions. The challenges addressed by AI belong to three different categories: maritime, port and hinterland. On the other hand, identified AI technologies are presented within four categories: machine learning, deep learning, evolutionary computation, and uncertainty/probabilistic handling. This research collects 17 different AI technologies and 47 distinct challenges.
The results show that most challenges are related to sea carriers, terminals, and nautical chain operators (towage and pilot companies). Terminal operators have the highest share of challenges, and within the terminal operator sector, quayside operation has the bigger share compared to landside. The research also reveals that Deep Learning (DL) has the highest capacity for addressing challenges identified in this research.
Moreover, we see also that almost all AI technology methods can contribute to tackling “(Autonomus) ship collision avoidance” challenge. This can help stakeholders and technology providers realize the potential of AI for this challenge. All DL methods can address the “(Autonomus) ship collision avoidance” challenge. In contrast, only 66% ML methods can solve this challenge. This way, stakeholders would understand each AI method's capability in addressing each challenge. However, DL possesses the greatest proficiency among AI categories in handling these specific challenge domains. Furthermore, when comparing the capabilities of ML and DL, it becomes apparent that, in most challenges, DL exhibits a greater degree of applicability than ML.

Programme committee
Freight and Logistics

Topic
Technology and Artificial Intelligence