A two-joint arm reaching under a sensory delay. Watch where the controller thinks its hand is, where its senses say it is, and where it actually is.
Sensory signals about the body arrive late. A controller that knows how its own body responds to commands does not have to wait for them: it can take the last state its senses reported and run that state forward through the commands it has issued since. Feedback still matters, but its job changes — it corrects the movement in progress and recalibrates the model that does the steering.
Workspace baseline
where the hand iswhere the senses say it iswhere the model says it isintended pathforce on the hand
Sensory lag error
—
Model estimate error
—
Model matches world
—
Model expects
—
Hand speed true, and what has arrived
Sideways error, reach by reach reach 0
What the model has learned force vs. hand velocity
Run
Bodyloop gain too high for the delay the model cannot bridge