A very interesting issue was raised in the discussion during the World-Cargo News industry leaders’ debate at the TOC Europe conference in Rotterdam in June. In considering the industry’s drive to digitalisation, the question was posed as to whether container terminals could benefit from artificial intelligence (AI), and in particular whether ‘intelligent’ machines could reduce the number of exceptions that slow down automated terminals, by recognising problems and making decisions without the need for manual intervention.

Representatives from Kalmar, Konecranes, ABB and Siemens all saw AI as playing a major role in terminal automation in the future, but Dr Yvo Saanen, TBA commercial director, presented a dissenting voice, arguing that trying to empower machines to make decisions for themselves is not necessarily going to solve the problem that terminal operators are trying to address.

The difference of opinion really gets to the heart of terminal automation, where, for several years now, the focus has been on making the terminal process predictable and repeatable, as far as possible, to the point where container handling can be automated.

Making progress

A lot of progress has been made, but the automated container terminal is not yet on the same level as an automated factory line, in terms of predictability and performance – the process stops repeatedly for systems to make decisions, and to solve exceptions. Automated cranes move quickly and efficiently when they have a job to do, but then they seem to stop and wait.

At the same time, the development of AI is driving the development of more intelligent robots, theoretically able to do much more than just repetitive tasks on a production line. By leveraging machine learning and deep learning, robots will be able discover patterns in data, and perform better, without having to follow explicitly programmed instructions. In a container terminal context, machines could, possibly, resolve a lot of exceptions on their own.

AI is exciting and one of the leading topics in robotics and automation today. Saanen said there is a fair bit of hype around what AI actually is, however, and how it might change a process like terminal automation. “We are talking in the end about decision algorithms, located somewhere in the system, whether they are in the cloud or on the machine,” he said. “AI brings a lot more data about the operating environment into that process, but it is still a data-based process.”

In some applications, the ability to use vast amounts of data could give autonomous machines an advantage over human decision makers, but Saanen makes the point that a container terminal environment is not like using big data sets to detect fraudulent patterns in financial transactions, for example. In container handling, the processes to be performed are actually quite narrow, and very discrete, “notwithstanding that pattern analysis can bring improvement to operations no doubt”, added Saanen.

One of the big misconceptions around AI in the container terminal environment, he continued, is that there is a lack of data that is somehow causing delays and exceptions that AI can help solve. In fact, the real problem is data quality – automation systems are working with data that is inaccurate, or is missing altogether, and frequently subject to change.

As an example, Saanen pointed to loading lists, which show the sequence in which containers should be handled. These are subject to continual change, for a variety of reasons, including containers being stowed out of sequence. This affects the whole automation system. “Here you are, with all your predictive logic and intelligence being based on wrong information,” Saanen noted.

Adding more levels

More and more levels of sensor data are being added all the time, to measure other factors like container weight and proximity between equipment inside the terminal, as well as data on the position of road trucks and other external events. But can all this extra data lead to a better container handling process within the terminal?

“I am not a disbeliever in technology,” said Saanen, “but terminals are very small in scope – there is very little benefit in having intelligence on a stacking crane or an AGV or terminal truck.” AI and machine learning has not changed TBA’s fundamental belief that the process of moving containers between different modes of transport in a closed terminal environment should be controlled from a centralised environment, by a system that makes decisions holistically.

Where TBA sees scope for a lot of improvement in terminal automation systems is improving data quality, and further development and usage of better decision making algorithms, at both the TOS and equipment control system (ECS) level.

Saanen believes that this is where the industry is falling behind. TBA’s TEAMS ECS has a lot of decision making logic that can actually be described as AI, “yet the focus is predominantly on execution, rather than optimisation”, he said. “We can do a lot more to improve decision making.”

The industry as a whole needs to address its data accuracy issues, but Saanen also acknowledged that TOS and ECS suppliers, including TBA, need to do a lot of work to improve their products to meet the needs of terminal operators. Saanen added that AI will not, however, be driving TBA in a new direction. At the end of the day, AI is really automated decision making logic in machines and systems. Those processes can be improved by using more data and processing it in a better way, but terminals still need to address the integration between the TOS and ECS, and the data issues that are problematic today, emphasised Saanen.

TBA continues to develop TEAMS, and is now working on functionality to support Ports of Auckland’s new straddle carrier automation system, which will use TEAMS, integrated with the Navis N4 TOS and Konecranes Terex straddle carriers. This is a different combination than has been deployed before, and TBA will be using around 80% of the modular functionality in TEAMS, with the remaining 20% of the requirements needing new development. TBA will also deploy its Scheduling and Dispatching module for both the manned and automated straddle carriers.

Autonomous trucks at the terminal gate

Autonomous trucks represent a difficult issue for container terminal automation. At terminals today, almost all automated machines are separated by a fence, with multiple levels of collision avoidance and vehicle detection systems. Autonomous road trucks are being developed to operate on public roads, and several companies are now offering automated terminal tractors.

Some of the leading terminal operators are keen to see this develop quickly, both as an alternative to more expensive AGVs, and as an option for automated horizontal transport in existing RTG terminals. TBA’s Yvo Saanen said this scaling down of automation is, “in general, a very good way forward”, and will lower the cost of horizontal transport automation for terminals.

Putting automated terminal tractors into a conventional RTG terminal, however, is still a considerable challenge. Whether the tractor is autonomous or sensory controlled, like an AGV is today, it will have to mix with manned machines, including road trucks. There will need to be a system for making sure machines do not collide, and that system will have to be certified in the same way that crane automation systems are today. At this stage, there is no clear indication of what this might require, or how it can be achieved.

There is some expectation that terminals will fall under new regulations being developed in the US and the EU to allow autonomous vehicles on public roads, but these have not been finalised at this point.

Read this item in full
This complete item is approximately 1000 words in length, and appeared in the September 2017 issue of WorldCargo News, on page 26. To access this issue download the PDF here.