Teach-and-Playback Programming

A human first guides the robot through a motion and records its positions, and the robot then repeats exactly what was recorded.

This is the dominant way traditional industrial robots are programmed: an operator uses a teach pendant, a handheld programming device with an emergency stop and an enabling switch, to jog the robot, or physically guides the arm by hand, recording a sequence of key positions and paths; afterward, the robot “plays back” this recording over and over. Unimate, the first industrial robot, which went into service at a General Motors plant in 1961, worked exactly this way, storing taught joint positions on a magnetic drum and replaying them. This method is simple, reliable, and highly repeatable, well suited to fixed-position repetitive work such as welding, spraying, and material handling — but the robot merely reproduces a trajectory without understanding the task, so it has to be re-taught whenever a workpiece's position changes. Imitation learning in embodied AI also starts from human demonstrations, but the difference is that it learns a policy that can adjust its actions based on observations, rather than mechanically replaying a fixed recording.

ExampleIn an auto-welding shop, an engineer uses a teach pendant to move the welding gun to each weld point on a car body one by one and saves the sequence; from then on, every car body that arrives is welded by the robot following that same recorded program.

Also called

Teach and Playback, Teach Pendant Programming

Related

Teach Pendant · Kinesthetic Teaching · Industrial Robot · Imitation Learning · Structured vs. Unstructured Environment · Offline Programming

Sources

Teach pendant - Wikipedia
Unimate - Wikipedia
Industrial robot - Wikipedia