Jan 2018

Shipping: The Intelligence of Automation


Business is currently going through the most significant disruption in decades. This transformation was driven mainly by technological changes - large amounts of data and its analytical analysis, additive manufacturing, 3D printing, Internet of Things (IoT), robotics and artificial intelligence (AI) ) - are described as the fourth industrial revolution.

By 2016, about 17.6 billion devices were connected to the Internet. It is estimated that this figure reaches 80 billion by 2025.

Technology has touched and revolutionized almost every aspect of human life and the way we work. The maritime transport industry has also been touched by these technological developments in its mechanical aspect, but also digital, and they will determine the future of maritime transport.

Artificial intelligence alone can be the most disruptive technology since the industrial revolution and is changing the way organizations interact with consumers.

There are many definitions for artificial intelligence, but the true meaning seems to change as technology evolves. While some definitions characterize AI as "a technological approach to enable machines to do what we think only human beings can do," and others only as "the simulation of human intelligence processes made by machines, especially in computer systems" .

AI is by necessity a broader concept and largely encompasses machine learning (ML), neural networks, cognitive computing, and many aspects of natural (human) language processing. AI is a key to intelligent machines and systems and allows machines to have autonomous behaviors where little or no human intervention is required.

There is little doubt that AI will eventually impact almost every component of our daily lives. It is used today in many of our daily routines without really noticing it, in smartphones, smart homes, shopping tips and cars (not just the self-employed).

AI is allowing machines to play other roles and perform tasks that we previously thought could be done by humans alone, solving and facing problems in new ways (see Figure 1).

Why should we use artificial intelligence? In looking at the problems we have today and in the capacities generated by AI, it has most of the capabilities we require:
• Self-learning and understanding of the reality in front of us;
• Adaptation when something is not as planned;
• Pattern prediction and recognition to be able to follow a path;
• Deep analysis and recommendations to the user.

The responsiveness in dynamic and complex environments has already been proven in the airline and banking industry, where the requirement is very high, full of dynamics and updates in real time.

Artificial intelligence is not just a technology, but a variety of different types of software that can be applied in various ways to different needs (see Figure 2).

AI does not think for itself. Your ideas are possible because the software is information-driven, and the more information you get, the more information it can produce.

People in your day to day carry huge amounts of data for your personal cloud. All this creates data that is available for the machines to use during the so-called machine learning process (ML). With technology in general and AI in particular, the key ingredient to success and value creation is having a first approach focused on people.

AI can empower people to create, think and innovate at totally unknown levels to drive growth and productivity.

Not only aiming for the elimination of repetitive tasks, AI should put people at the center, increasing their work capacity and applying machine capacity so that people can focus on analyzing data that brings more value, and which can help decision making and innovation in some processes (see Figure 3).


So, what is AI? How is it related to machine learning (ML)? And with the robot automation processes (Robotic Process Automation-RPA)? Will robots continue to "consume" our jobs in the future? What do we know about their learning methods and how we can work together with them?

AI is the intelligence displayed by machines / devices that allows them to evaluate factors in a given environment and perform actions that maximize their success around a given goal.

Not all processes can be automated. Processes that are characterized by the following criteria can be considered candidates for automation:
• large volume of manual work;
• rule-based / no human decisions;
• few exceptions;
• with digital drives and associated digital structures;
• repetitive and with few changes in procedures;
• high probability of human error.

The artificial intelligence (AI) technology market is booming. AI now includes a variety of technologies and tools, some already tested, others relatively new (see Figure 4).

Predictability must now be an integral part of the logistics chain. Immediately after the ship departs from a port, forecasts for the next ETA and the time spent on the next stopover in unloading and loading operations begin.

Each issue is affected by a decision made earlier on one or more links in the logistics chain. Each decision is based on the data of each participant in that chain.

In the past, industry gurus used to make these predictions based on their accumulated knowledge and sometimes with a few magic steps. Today it is impossible to work on this basis, with the number of movements made on each scale and the constraints faced in each port.

IoT is responsible for this incredible amount of data. From the weight of the containers and delivery dates and survey to the position of the ship and the weather, everything is processed almost in real time. These raw data are then cleaned, organized and processed, and this is the field of the machines in their learning (ML) and in the generation of intelligence.

Learning the machines is like the engine room of Artificial Intelligence and the data lakes are your oil. It is where data collection and analysis takes place that leads to the non-human decision-making processes that we call AI. It is possible to recognize certain patterns in the logistic chain that humans, per se, would hardly be able to detect. Machine learning is already present in our lives on a daily basis, using data from previous decisions to make predictions about future decisions.

The application of machine learning to the AIS (Automatic Identification System) signals of ships opens up the possibility of altering or reorganizing routes that may result in greater profits for shipowners or making decisions on some port calls as a result of the loads.

Why has it taken some time to realize the potential of automated terminals in operational performance? And, for those who implement automation solutions, why is such a complex and difficult process? In the "promised land" of terminal automation, there is still a lack of confidence in how robots, software, and humans acting as a whole can achieve higher levels of operational performance.

Automated decision making is a cornerstone of this "promised land."

Automated or not, decisions at container terminals should be standardized, looking for the best way to deal with situations that may arise from a particular planning. They must also be resilient, adapting with the necessary flexibility in the execution of tasks.

So far, container terminals have encountered challenges to make automated decision making effective, with gains in efficiency, safety and sustainability.

In ports, there are three areas where using these technologies can increase productivity and performance. Dock porches, which carry out ship loading and unloading operations, park equipment in the storage area and containers carrying containers horizontally. These equipment can be operated by human means or by software. The planning side (park and ships) has to deal with many restrictions related to storage policies and restrictions on the side of ships and container handling equipment. It is now almost impossible for humans to do all these tasks without great technological support.

Collaboration platforms, such as NAVIS's XVELA, will provide and improve the visibility and reliability of information related to ship loading and unloading plans as well as greater connectivity between terminal operators and shipowners. Terminals will be able to access these plans at any point during the rotation of the ships, allowing the alignment of resources and the resolution of potential problems well in advance.

"Intelligent equipment" is now the mantra of global container terminal operators and they are responsible for flooding the data lakes, which will then be used in the decision management process.

Given the progress made by emerging technologies such as artificial intelligence and machine learning, it is natural to think that this will help optimization algorithms read the operational situation and then help apply fast and robust solutions into a reality full of its own dynamics and exceptions.

We will continue to have robots, manpower and software in most automated container terminals. We need these actors to continue to do what they are good at and to help complement the capacities of others to deal with operational reality.
• Humans are good at: • intuition driven by perception • multidimensional assessment; - creative and experience-based decisions.
• Robots are good at: - predictability; ?- precision; ?- repeatability.
• Finally, the software is good at: - Abstract intelligence and problem solving. But software can only act with clear, defined goals and reliable data that can help with decision making. This means that there are still some corrections to be made in:? - data analysis and correlation; ? - methods of learning the most demanding machines and in cognitive training.

When thinking about automated terminal decisions, software should be the bridge for humans and robots if they can help each other.
To ensure that the software can make this connection, we need to consider two main things:
• information still needs to be transformed into meaningful data. Many data still need to be extracted, analyzed and understood;
• Improve the visibility of information in real time to make better decisions and create relationships of trust between robots, software and users, fundamental for the growth and maturation of these relationships.


The vision of this article (Figure 5) is about a future transformed by artificial intelligence (AI) and machine learning processes (ML) as follows:
• On the robot side, DataBOTS will handle huge amounts of data and will focus on repetitive automated tasks to find new relationships and transform data into meaningful information;
• Artificial intelligence can help us by managing unstructured data and performing complex tasks;
• On the software side, more developed algorithms are able to produce more knowledge and structure the data in a reusable way and with new configurations. But artificial intelligence and machine learning can help systems be more effective: • AI can analyze and correlate data, including generating new data • ML can constantly train and update data by automatically adjusting configuration of the parameters that control the behavior of the algorithms;
• For the user, UIBOTS will help process results and provide real-time feedback to drive proactive user interaction and effective exception handling, in which case: - AI can mimic the human brain and thus drive a reaction to new situations that may arise; - ML can integrate cognition and better behaviors in the performance of machines.
All these different elements will work together to have a major impact on short- and long-term planning, and flexible implementation of procedures.