Shield AI and Mitsubishi Heavy Industries (MHI) are pushing to make drones more autonomous, enabling them to sense, adapt, and act independently while tracking targets and coordinating with one another.
To do this, the companies paired Shield AI’s Hivemind autonomy software with MHI’s Affordable Rapid-Prototyping Mitsubishi-Drone (ARMD) in a recent flight test in Japan.
During the demo, two drones autonomously executed reinforcement learning, preprogrammed behaviors, and coordinated maneuvers while tracking a virtual air vehicle.
In the first flight, Hivemind used reinforcement learning to guide an ARMD in tracking a virtual target without human input.
In the second, the system directed another drone to carry out more aggressive maneuvers while pursuing the same type of target, highlighting more advanced autonomy behaviors.
“Our collaboration ability to transform how autonomous capabilities are designed, integrated, and deployed at scale, enabling operational platforms to support Japan’s evolving defense needs,” said Nathan Michael, chief technology officer at Shield AI.
“Flight concepts like this are the future of autonomous airpower, and each flight test marks one step closer to operationalizing at scale.”
Compressed Development Cycle
The demonstration also highlights how quickly advanced autonomy can move from concept to flight.
Working together, Shield AI and MHI engineers progressed from software testing to hardware validation in under two weeks, compressing a process that previously took months.
They then moved to live flight testing in another two weeks, significantly shortening the overall development cycle for mission autonomy systems.
