Technology

Artificial Intelligence: Ports Are Beginning To Take Up Positions

Artificial intelligence (AI) is an affordable technology, although it is only slowly being introduced into the business sector. Thus far, it has primarily been used to improve sales prediction techniques, but its potential applications are infinite and include lowering maintenance costs, improving product quality, planning manufacturing and increasing service level. In the sphere of transport, AI opens up a host of possibilities. Will the ports take advantage of them?

Posted on 09.20.2018

by Lluís Miró

Lluís Miró is a consultant for the Logistics and Ports Department at IDOM where he has served as Project Director and expert in port logistics. He has directed projects in the main Spanish ports, as well as in countries such as Chile, Mexico, Uruguay, Egypt or the United Arab Emirates.

Big data and IA can lead to a predictive, more efficient supply chain [Image by Mika Baumeister]

Today, the ports of Hamburg, Rotterdam and Singapore have already started to develop AI tools to improve predictions of maritime and land transport operations. Specifically, Hamburg has created a decision-making support system based on a predictive model of the behaviour of land transport. The model takes historical data, and using deep learning techniques and neural networks, it offers detailed predictions of the times when lorries should reach terminals. Based on this, the system notifies the lorry drivers of the terminal entrance times, and it gives the terminals a dynamic forecast of the workload they will have according to the changes in the surrounding conditions (road and access route saturation, real ship arrival time, degree of terminal saturation, etc.).

How does deep learning work?

Deep learning and neural networks are two of the machine-learning methods which have come to the fore the most in recent years. They are inspired by the way neural networks work in the brain. They transform the entry values, layer by layer, until the value of the variables that they are trying to predict is found. Even though the results of neural networks are quite satisfactory, they need vast amounts of data to learn, and learning times are long (days or even weeks). Natural language processing, image pattern recognition and voice processing are the main success stories of deep learning. Thus, the evolution of data collection and management has to include the following levels: recording, analysing, simulating, predicting and finally recommending. Based on that, new-generation ports are expected to apply predictive and prescriptive analysis techniques as tools to support decision-making when planning the transport of the actors in the port-logistics chain. And this does not only include lorries, since the same transport logistics that it applied on motorways can also be applied to any means of transport (railway, maritime or river).

The digital transformation in the port and the logistics chain entails huge amounts of data, many of them in real time. The competitiveness of future ports will largely depend on their ability to make use of this information. With AI tools that enable them to take advantage of the potentiality of this vast trove of data, the decisions taken by the managers will be higher quality, shared and generated more quickly, so they will likely optimise the time, cost and reliability of the operations in port-logistics environments. In a complementary fashion, all of this will end up leading to more flexible, real-time operations management. AI has reached the world of transport, and it is here to stay. The ports which realised its benefits and potentiality to change the sector first will unquestionably see operational efficiency gains compared to their competitors. Ports that already have advanced systems that allow them to gather a significant amount of data (Port Community Systems, Port Management Systems and Terminal Operating Systems, among other systems) will be the best poised to successfully incorporate the tools offered by artificial intelligence.

The origins of artificial intelligence

Even though it seems like a recent concept, the origins of artificial intelligence date back to the Greeks. Aristotle (384-322 BC) was the first to determine a set of rules that partly describes the way the mind works to reach rational conclusions, and Ctesibius of Alexandria (285-222 BC) built the first self-controlled machine, a water-flow regulator (rational, but without the ability to reason). John McCarthy, Marvin Minsky and Claude Shannon coined the term artificial intelligence at the Dartmouth Workshop (USA) in 1956 to refer to the “science and inventiveness of making intelligent machines, especially intelligent calculation programmes”. Where these three scientists missed the mark was in their prediction of when the first smart machines would arrive. They trusted that by the 1970s we would be surrounded by artificial intelligence. However, the majority of tech companies did not decide to make significant investments in this field until the 1990s and 2000s, in a bid to improve the processing and analytical capacity of the vast amounts of data which were being generated in the new digital world. In fact, AI was definitively enshrined in 1997, when IBM demonstrated that an IT system was capable of beating a human at chess. And it wasn’t just any human; it was the world champion, Garry Kasparov. The supercomputer was called Deep Blue, and it marked the turning point when industrial technology and society at large became aware of the real importance and possibilities of artificial intelligence.