When natural disasters destroy infrastructure, delivering emergency aid becomes a massive logistical challenge. To address this, humanitarian relief teams often rely on a “tag team” approach that uses trucks and drones.
Now, researchers at the Stevens Institute of Technology have developed a new artificial intelligence algorithm that optimises this truck-drone delivery system to prioritise fairness alongside efficiency.
The study, published in the journal Computers & Industrial Engineering, outlines a mathematical model that shifts the focus from simply reducing average delivery times to minimising the time it takes for the very last person to receive assistance.
Truck and drone tag team
In disaster scenarios, trucks serve as mobile headquarters, hauling large amounts of food and water as close to the affected site as possible. When roads become flooded or impassable, drones take over the “last-mile delivery” by flying over obstructions to drop off urgent supplies.
Stevens Associate Professor Jose Ramirez-Marquez explained the benefit: “This ensures that aid reaches even the most isolated and inaccessible locations. The drone can come back, and then it can be resupplied, and then deliver aid over and over again.”
Working alongside Teaching Assistant Professor Nafiseh Ghorbani-Renani and PhD candidate Ramin Talebi Khameneh, Ramirez-Marquez used an evolutionary algorithm to generate optimal delivery routes.
Instead of just aiming for speed, the bi-objective model actively shortens the time difference between the earliest and latest deliveries while also reducing the total distance travelled.
“That way, aid is spread more evenly over time,” Ramirez-Marquez noted. “And it’s also a fairer way to deliver aid.”
To prove the system works, the team ran simulations based on two distinct disaster scenarios:
- Urban flooding: A simulation based in Hoboken, New Jersey, a city with a history of severe inundation following Hurricane Sandy in 2012.
- Rural flooding: A simulation based on the 2025 flash floods in Hopkins County, Kentucky, a region filled with low-lying roadways.
During the rural simulation, the researchers also introduced the threat of disinformation, which can lead to misleading aid requests. The algorithm evaluated how different prioritisation strategies can safeguard equitable access to resources when post-disaster information is incorrect or unverified.
The research team says the algorithm is now ready for deployment and is seeking a partner municipality to conduct a real-world test run.
