Kenneth Glad
When hunger strikes after a long day, the wait for dinner can feel endless. You watch the clock and hope your meal arrives before the heat escapes the box. Soon, the math behind that journey might look very different.
A research team in Denmark wants to fix takeaway logistics, so they’ve built a new artificial intelligence model that aims to get food to your door with less wasted effort.
It moves away from simple speed or lowest cost. Instead, the system looks at the bigger picture of how drivers move around a city.
“The crucial thing is that the model does not only try to make deliveries as fast or cheap as possible,” Yihua Wang from the University of Southern Denmark told TV 2.
“It weighs several competing considerations and finds a balance between efficient routes, timely deliveries and the available courier capacity,” she explained.
Tested on massive scale
The team published their findings in the journal Transportation Science. They ran the numbers on a massive scale to prove the concept works.
They analyzed 650,000 orders from Meituan, a major delivery platform in China. The results showed a clear shift in efficiency.
The model required up to six percent fewer couriers per order. Yet it still managed to deliver more than 95 percent of those meals exactly on schedule.
The AI also spotted clever ways to bundle trips. It grouped up to 24 percent of orders together, which beat the platform’s standard practice by a solid margin.
Looking for local partners
The technology exists today, but no delivery app uses it just yet. The team hopes to change that soon.
“The next step will be close collaboration with a delivery platform so the model can be integrated into its system and tested under real operating conditions,” Wang said.
She specifically pointed to a certain major player in the Danish takeaway market: “It would certainly be very interesting for us to collaborate with Wolt.”
An open door
The delivery company seems open to a partnership. Jacob Schmidt Bollerup, a regional manager for Wolt, addressed the project:
“We are always interested in collaborating with initiatives that can make us sharper and better.”
He noted that Wolt already uses algorithms to manage its drivers. Even so, the university researchers believe their model offers something totally new.
“There is still a very large potential for development, and our model is unique,” Wang told TV 2.
Sources: TV 2, Transportation Science
