In industry, several requirements often need to be met at once. A product may need to be affordable, sustainable and of high quality, while also being efficient to produce. Improving one aspect can sometimes come at the expense of another. In his doctoral thesis, Anan Ashrabi Ananno, a doctoral student at the Division of Product Realisation, has explored how optimisation can help identify and compare possible solutions when several goals and constraints need to be considered.
"I often say that I teach computers to make decisions about complex problems. Mathematics doesn't care whether it's dealing with a box or a meatball. The underlying principles of optimisation can be the same," says Anan Ashrabi Ananno.
Robots that stack boxes on pallets
When stacking goods on a pallet, making the most of the available space is only part of the challenge. The boxes must also be stable and able to withstand the weight of other goods. Similar products may need to be placed close together to make them easier for customers to handle. And if a robot is doing the stacking, it must be able to reach and position each box without colliding with anything already on the pallet.
The research therefore combines optimisation with a digital model of the robot. The model is used during the planning stage to check whether the robot can reach and position each box. This allows solutions that the robot cannot carry out to be ruled out before the pallet is built.
The method has been tested both digitally and using physical robots. During testing at the company OpiFlex, 150 trials involving different orders were carried out, all of them successfully.
Automation could also improve working conditions. Stacking goods on pallets by hand is physically demanding and repetitive work.
"Building pallets by hand is hard physical work. Doing it for 20 or 30 years can lead to chronic health problems. That's why we wanted to see whether a robot could take over the task," says Anan Ashrabi Ananno.
From boxes to meatballs
During his doctoral studies, however, the research took an unexpected turn. Anan Ashrabi Ananno began exploring whether the approach used to solve pallet-stacking problems could also be applied to food product development.
Working with researchers in culinary science, he used meatballs as a case study. Instead of deciding where to place boxes, the challenge was to work out which ingredients to include in a recipe and in what proportions. Several factors, including taste, nutritional value, cost and production time, had to be considered at the same time.
One question was how a recipe could be adapted if there was a shortage of meat and other protein sources had to be used instead. The food would still need to be nutritious and taste good.
鈥淚n a crisis, we can't assume that all ingredients will always be available. But the food still needs to taste good. That's especially important when serving children, who can be sensitive to changes in flavour,鈥� says Anan Ashrabi Ananno.
More than one solution can work
One important finding is that optimisation does not necessarily have to produce a single answer.
When stacking pallets, making the most of the available space may conflict with what a robot can physically do or how the pallet will later be handled in a warehouse. When developing a recipe, factors such as taste, nutritional value, cost and production requirements need to be balanced.
Optimisation can therefore help identify several possible solutions, each with its own advantages and disadvantages. People can then decide which compromises make the most sense for their situation. Rather than aiming for one predetermined optimal solution, the thesis explores how different alternatives can be generated and compared.
"You can't get the best result in every category. You have to decide which trade-offs you're willing to make," says Anan Ashrabi Ananno.
The same approach to different problems
This is where the connection between pallets and meatballs becomes clear.
The thesis does not suggest that exactly the same mathematical model can be used unchanged for both robots and recipes. Instead, it shows how a similar approach can be applied to very different problems: creating a digital representation of what is being designed, identifying constraints, evaluating several objectives and allowing decision-makers to compare alternatives. The specific models, variables and constraints must still be adapted to each application.
This was one of the things that surprised Anan Ashrabi Ananno during his doctoral studies.
"Mathematical models don't know whether they're dealing with food or a box. They just work with numbers. With the right mathematics, you can use a similar approach to solve problems in completely different fields," says Anan Ashrabi Ananno.
