Give the arm an instruction in plain language. PollenX turns it into a plan and, before every move, states what the sensors should read. The workcell then measures whether they did, and recovers when they didn’t. Everything runs live in this tab.
What is real here. Every run is simulated live in this tab: the arm, the contact physics, the load cells, the wrist force sensor and every check in the ledger. Vision is deliberately imperfect, the way a real camera is, so the same instruction can play out differently twice.
How the plan is made. PollenX reads your instruction against what the workcell can see right now, breaks it into steps the arm can carry out, and attaches to each step the measurement that will prove it worked. Then it steps back. The arm acts, the sensors report, and the checks decide, so PollenX never grades its own work. If an instruction can’t be done with what is on the bench, PollenX says so and nothing runs.
When a step falls short. The arm retries. When retries are spent, the system re-observes where everything actually ended up and re-plans for what is still undone. After two re-plans it stops and hands the measurements to an operator.