3 Video Views·Mar 19, 2026
Shield AI successfully completed autonomous flight tests outside Tokyo, Japan in collaboration with Mitsubishi Heavy Industries, Ltd. (MHI), integrating Hivemind autonomy software onto MHI’s Affordable Rapid‑Prototyping Mitsubishi‑Drone initiative (ARMD) in under 60 days. The demonstration underscores how advanced autonomy can be rapidly developed, integrated, tested, and flown in support of defense-focused autonomous air operations.
The flight test, conducted in Ota City, Gunma Prefecture, culminated in two live flight demonstrations, during which two different test aircraft autonomously demonstrated reinforcement learning, trained behaviors, and coordinated motions while tracking a virtual air vehicle. During the first flight, Hivemind used reinforcement learning to autonomously direct the first ARMD in pursuing a virtual object. Hivemind executed more aggressive behaviors to pursue the virtual object during the second flight test, exhibiting advanced reinforcement learning motions on the second ARMD.
MHI has been developing mission autonomy using its in-house AI development environment for over a decade
We believe it is crucial to develop mission autonomy domestically to determine the future of Japan's drone operations
This unit is the ARMD specifically the ARMD Model 1 and Model
2 These units were developed by Mitsubishi Heavy Industries
Equipped with various AI systems it serves as a flight testbed capable of flight testing
Building and maintaining an AI development environment requires significant effort
On the other hand leveraging HybriMind Enterprise as an AI development environment allows
MHI can focus on developing mission autonomy
This demonstration has enabled Japanese mission autonomy to be achieved within eight weeks
We successfully demonstrated the ability to develop test and fly within a short timeframe
In this project we utilized a hive mind to conduct testing entirely within simulations
We advanced the development seamlessly all the way through testing with actual hardware
What moved me most was how quickly they developed it implemented AI launched it and actually conducted flight tests
This white aircraft has its waypoints generated by our company's developed AI
This would be equivalent to flying the aircraft through the waypoint using the autopilot
This white aircraft is flying while following a virtual target aircraft in flight
This aircraft is operated by an AI that generates control commands to pilot the vehicle
It is an aircraft that grants greater authority to the AI to enable aggressive flight maneuvers
Hibumine's enterprise significantly accelerates the development of Japan-made missions by MHI
Over the past seven or eight weeks I've been working hard alongside the HiveMine project
Working on my HiveMine project together with it
I believe we can succeed in future
projects Domestic support is also necessary