Depending on who you ask, package delivery by drone is either looming large in our future or is way off in the distance. While we wait, researchers at MIT are building the parts that will make delivery robots functional and efficient, including battery-swapping stations and algorithms that streamline the way the crafts will think.
Calculating a flight path to a delivery destination while staying fueled and bracing for air currents takes a lot of computing power, which is typically limited on a craft like a quadcopter.
A new approach to computing these scenarios separates the various calculations, in this case location and “health”—whether a blade on the rotor is broken, or if fuel is running low—Ali-akbar Agha-mohammadi, a post-doc at MIT’s Laboratory for Information and Decision Systems, explained.
“We have broken the large problem into lots of small problems,” Agha-mohammadi said, and as a result they’ve decreased the computing power needed.
Agha-mohammadi and his colleagues are also engineering hardware. They’ve built charging stations that a quadcopter can land on to recharge or replace a battery, and electromagnets that can pick up packages.
The new algorithm has proved its effectiveness in a simulated drone system, and Agha-mohammadi will present the results at an upcoming conference in Chicago. He says the same underlying principle could apply to robots on the ground as well.
The next step he says is to get the quadcopters to test the complete takeoff-recharge-deliver sequence in the lab.