Count the actuators and the story changes; count the reaction forces and it changes more. US10974830B2, granted to Aurora Flight Sciences in April 2021, mounts a robotic manipulator on an aircraft — and in doing so makes visible a constraint that every robot has but most can ignore.
The B25J filing tells a different story than the keynote. Classified under B25J 9/1065 and B25J 15/0213 for the manipulator and B64C 39/024 for the aerial platform, the patent has to confront Newton's third law in a way a floor-bolted arm never does. When a grounded robot grips and pulls, the floor absorbs the reaction. When an airborne robot does the same, the reaction destabilizes the aircraft itself.
This is the unstated tax on every manipulation claim: a grasp is not just about the hand, it is about whatever the hand is attached to staying put. Industrial arms get to assume a rigid base. Humanoids do not — a robot reaching out shifts its own center of mass — and aerial manipulators face the extreme version, where the base is actively trying to fly away.
The honest engineering the patent implies is coupled control: the manipulator and the flight controller cannot be designed independently, because each one's actions are disturbances to the other. That coupling is exactly the problem humanoid balance faces when an arm reaches, just dramatized by gravity and rotors.
Aurora's filing is a useful teaching artifact precisely because it is extreme. It forces into the open the question every mobile manipulator must answer: where do the reaction forces go? A warehouse arm answers 'into the floor.' A humanoid answers 'into my own balance controller.' A flying arm answers 'into my attitude, so I'd better plan for it.'
For readers auditing the next humanoid that reaches out and grabs something on stage, the patent offers the right skeptical question: not 'did the hand close,' but 'did the body stay balanced while it did.' The reaction force is the part the demo lighting hides.