DARPA looking to give UAVs better 'hand-eye' coordination

V-Bat extending arm

Researchers funded by the Defense Department are hoping to give unmanned aerial vehicles finer visual depth perception and capabilities that enable them to pick materials up and move them around.

The Defense Advanced Research Projects Agency (DARPA) on Dec. 6 detailed some of the research being conducted to give unmanned aerial vehicles (UAVs) more precise autonomous payload placement capabilities. 

Researchers funded by the agency successfully tested a vision-driven robotic-arm payload emplacement using Santa Clara, CA-based MLB Company’s V-Bat “tail-sitter” UAV. The vehicle, said DARPA, is capable of both hover and wing-borne flight, making possible the delivery and precision emplacement of a payload.

Key to the capabilities is a special extendable six-foot robotic arm attached to the UAV that can grab and carry up to a one pound payload.

The research team designed and developed a low-cost vision system that estimates the target’s position relative to the hovering vehicle in real time, said DARPA. This vision system also enables the UAV to search and find the target for the emplacement autonomously and then perform the action, it said.

The capability paves the way for a number of applications for precise long-range delivery of small payloads into difficult-to-reach environments, according to the agency.

“Our goal with the UAV payload emplacement demonstration was to show we could quickly develop and integrate the right technology to make this work,” said Dan Patt, DARPA program manager. “The success of the demonstration further enables the capabilities of future autonomous aerial vehicles.”

During the technology demonstration, DARPA said the MLB-built V-Bat successfully demonstrated:

• A newly developed stereo vision system that tracks the emplacement target and motion of the robotic arm. The vision system, coupled with global positioning system, controls the arm and V-Bat during emplacement.
• Control logic to maneuver the vehicle and direct the robotic arm to accurately engage the emplacement target.
• Vehicle stability with the arm extended six feet with a one-pound payload.
• Autonomous search and detection of the emplacement target and autonomously emplaced a one-pound payload.