A growing interest in using artificial intelligence to gather and analyze logistics data is raising the possibility that shippers, freight forwarders, and carriers can make vastly more informed planning decisions to help optimize the travel time, cost, and resources needed for moving cargo.
Artificial intelligence (AI) advancements in gathering and analyzing logistics data are providing ways for the container shipping industry to plan further out and more accurately. AI — a science-fiction trope increasingly seen as being the next digital game changer in a variety of industries — could help containerized supply chains better ensure just-in-time transits and equipment availability — two major cost pinches for shippers.
Companies including Maersk, Panalpina, and Flexport are trying to harness the techniques of AI — loosely described as the use of computers to simulate human intelligence — to tackle a variety of vexing industry problems, with the help of IT companies such as ClearMetal and Maana. They are targeting issues such as how to pick the best alternative port when the original destination is blocked, better estimating the arrival time of a ship so that logistical resources can be ready. AI is also being tapped to forecast whether a shipper will cancel a booking or its container will get rolled by the carrier, and left on the dock.
So far, the use of artificial intelligence to solve such problems is in the early stages. Additionally, the advance of such techniques could be slow, given the industry’s past reluctance to embrace other types of technology.
However, industry officials said the ability of artificial intelligence techniques to analyze far more data than can be done with traditional methods, and the fact that vast amounts of data is already available in the shipping and logistics industry, mean it is ripe for the use of AI techniques, offering the potential to dramatically curb unpredictably and improve visibility in certain parts of the supply chain.
“In a world as complex as supply chains, with so many interdependencies between variables, there should be fertile ground for artificial intelligence,” said Ryan Petersen, CEO of Flexport, a San Francisco-based digital freight forwarder startup that is working to incorporate artificial intelligence techniques into its products. “Shippers will benefit from better decisions about when and how their cargo is shipped to lower working capital needs, transit times, or logistics spend, depending on their preferences.”
Most commonly associated in popular culture with fictional robots that talk and respond like humans, the term artificial intelligence covers a wide range of digital capabilities, from the design and creation of robots and driverless vehicles to computer vision, or the extraction of data from an image.
In the logistics and shipping context, artificial intelligence is currently most focused on large-scale number crunching that can analyze and organize data from different sources, shape it, and then use it as the basis for decision-making, sometimes with reduced or no human input.
That ability is enhanced with “machine learning,” a subset of artificial intelligence, in which algorithms analyze data and, based on the knowledge gleaned in the results, adjust their logic on an ongoing basis to provide a more accurate analysis.
Machine learning, for example, is at the heart of a mathematical model developed by ManWo Ng, assistant professor of maritime and supply chain management at Old Dominion University, to project a port terminal’s future chassis requirements.
Ng’s model analyzes factors such as how many vessels are at berth, the number of import containers discharged on certain days, how many export or empty containers were received, and how many gate transactions were recorded. Some of these are the same factors taken into consideration by chassis operators for their own predictions. However, Ng believes that the constant refinement of machine learning produces an algorithm able to far more accurately predict future chassis demand.
In one use of his program to study of a major US marine terminal, Ng concluded that the terminal was grossly overstating its chassis requirements, and could cut its chassis repositioning requests by 80 percent, the study found.
“Thus, not only can chassis demand be better predicted, the availability of predictive models also means lower repositioning costs and a reduced environmental impact due to the elimination of unnecessary chassis repositioning trips,” Ng said.