Artificial intelligence: When machines learn how to learn
Oct 20, 2017 Comments (0)
Two pairs of socks, a toy car, and a jigsaw puzzle come down the conveyor belt.
The employee in the logistics centre grabs a box, but can’t fit all the ordered items into it. So he chooses a new box and tries again. What used to be a common scenario at packing stations is now a thing of the past, thanks to artificial intelligence. Today it’s no longer humans who decide which box is the right size. Intelligent software is continually gathering feedback from customers and employees, and combines it with product data. The software recognises patterns in this data and uses them to choose the right packaging dimensions. This not only saves time, it also saves packaging materials and shipping costs.
But when is a system actually considered intelligent? In a nutshell, machines are considered intelligent when they learn something that was not precisely pre-programmed in their code. In other words, we provide machines with tools that they can then use to shape their own environment. “Training rather than programming” is the principle behind AI.
In logistics, AI is just starting to gain ground. That’s because the challenges are now becoming too complex for traditional methods. Tasks and processes for comprehensive planning and control seem particularly predestined for AI.
Perhaps the best-known example in the area of planning is the forecasting of sales and demand. Such forecasts are difficult to obtain and often deliver disappointing and inaccurate results. AI can make life easier by providing reliable models to predict variability in demand. It does so by collecting all demand-related attributes, such as activities on the company’s website, while filtering for “disruptions”, such as random and unforeseeable fluctuations in demand. The system learns from the data and adjusts its model to changing consumer behaviour.
While research on predictive analytics is still in its infancy, it will eventually be possible to generate much more precise and efficient algorithm-based forecasting. That’s why businesses should start determining today whether their own supply chain can be optimised by intelligent demand forecasting.
AI also controls complex situations where industrial robots are deployed. While these were long considered inadequate for the diverse tasks in logistics, the current generation of robots can see, move, respond to their environment, and perform high-precision tasks alongside human beings.
Take the Toru Cube, a warehouse robot that is the brainchild of the start-up Magazino. This robot receives orders by Wi-Fi and finds its own way around the warehouse using laser sensors. When it reaches the designated shelf, it uses cameras and lasers to compare the dimensions it sees with the information stored in the database. Once it “knows” it has found the right item, it can use its robotic arm to retrieve it. While the Toru Cube moves through the warehouse much slower than a human being (for safety reasons), it can work three shifts without a break and doesn’t make picking errors.
Smart robots are also an attractive option for last-mile delivery. The Starship robot in Hamburg, for example, is equipped with cameras, GPS, and gyroscopes for navigation. It rolls through city centres at 6.4 km per hour, but isn’t totally self-reliant yet. If it needs to navigate among too many people, it can summon an operator to take over the controls. That’s particularly attractive for logistics service providers, as the last mile is relatively expensive.
If many companies are still reluctant to use AI in logistics, it’s not because they lack expertise, but because logistics isn’t considered the priority it should be. And yet, its importance can hardly be overstated, seeing that it may well be the competitive factor of the future, particularly in logistics. Where companies used to differentiate themselves through mechanical automation solutions or process expertise, tomorrow’s business leaders will be those who best implement their digital strategy, harnessing their data to offer added value to their customers.
Darren Travers is a Senior Account Manager at AEB, responsible for key accounts in the domestic and international markets.