Would you take targeting orders from an autonomous artillery shell? That's the future the Army imagined in 1979.
A patent filed in that year and awarded in 1981 detailed an artillery round that would be fired towards a target area and then deploy a parachute. Then, it would slowly descend to the battlefield, taking pictures or video and identifying targets below. It would then feed the images and target positions to artillery batteries so the targets could be killed.
Figure: US Army patent application
That's right, the artillery shells would've been feeding targets to the gun bunnies.
This would've reduced the need to put artillery observers into harm's way when fighting against massed enemies. Instead of sending out a maneuver force or aerial reconnaissance patrol to find the enemy and feed targeting information back, the Army could just fire some rounds out there.
The system did include a "man-in-the-loop" function meaning that, like modern drones, a human would make the final decision on which targets would be killed. A crew chief would sit in a targeting van with a light-sensitive computer display. As the drone's imagery and proposed targets came up on the screen, this chief could designate new targets or remove target designations as necessary with a light pen.
The patent author specifically noted the importance of the chief completing this task since most computer systems of the day were prone to identifying large rocks and bushes as targets. Also, the remains of a destroyed tank still look very tank-like and could cause the computers on the artillery rounds to keep designating an already dead target.
Note the rock that is designated as a target. It's the human's job to go, "Lol, computer. No. Kill this other thing." (Figure: US Army patent application)
Modern battlefields contain more collateral damage concerns than many people envisioned during the Cold War, so this man-in-the-loop would also be useful as a final check to make sure a family SUV isn't targeted.
Once the computer had its final list of targets, more camera rounds would be fired at moving targets. These would contain explosive canisters instead of parachutes and antennas. The rounds would identify their designated targets, predict where the vehicles would be at the end of the rounds' flight, and then steer themselves to their final impact points.
Fixed targets identified by the system could be engaged by standard artillery rounds. Each round's impact point would be relayed to the firing artillery battery so that gunners could adjust their firing solutions if they missed.
(Figure: US Army patent application)
The patent also mentions the possibility of using a similar technique with helicopters. In that case, missiles would be used instead of artillery rounds and the human in the loop would ride in the helicopter, disapproving or adding targets to the computer from there.
Also, in place of the first missile being used to photograph or film the battlefield, the helicopter could pop up from behind cover to grab the first image.
In the end, there's no evidence that the rounds were ever completed. The Army had already experimented with placing cameras in artillery rounds in the 1970s, but that project was canceled due to technical problems. The patent for the autonomous system was filed in 1979 after the earlier program was already shut down.
The Army's plan to use aerial drones to target artillery lived on, though. Before drones were armed, they would designate targets for artillery or cruise missile strikes, a trick they can still do when necessary. In civil wars like those in Ukraine and Syria, both sides have used drones to spot targets for their artillery batteries.
Today, it's the Navy that's leading the charge for firing drones out of cannons. Their LOCUST program calls for dozens of drones to work together to canvas and attack a target. While other drones could be fitted into the program later, the ones currently being used are fired out of air cannons before spreading their wings and flying to their target.