A robot is the easiest part of a robot cell. The hard parts are everything it has to agree with — the conveyor, the gripper, the safety circuit, and a product that varies more than anyone admitted at quote.
Cells and arms
01Integrated cell — arm, infeed, fencing, light curtain. The fence line is a design decision, not an afterthought.02FANUC arm mid-integration. Yellow is not a coincidence; it is a whole ecosystem you commit to.03Collaborative arm in the shop. Different safety philosophy, different speed budget, entirely different conversation with the customer.04Cobot on a mobile cart — deployable to a station rather than fixed to one. Flexible, and a fresh set of problems.05Teaching and programming. Most of robotics is this: a laptop, a pendant, and a lot of small corrections.06The same cell, modeled first. Reach envelope proven before steel is cut.
On the floor
The photographs above are cells. This is what a plant full of them looks like — several running at once, sharing infeed, sharing aisles, sharing the operators who have to walk between them.
07Production hall. Arms, conveyor, gantry, and the yellow guarding that separates the two kinds of space.08Cells in a row, each one an independent machine that still has to agree with its neighbors about timing.
09Same line from the other end. The green conveyor is the thread everything else is hung on.10Gantry over the pack-out. Product stacked to the ceiling because that is what running well looks like.11Arm at rest between cycles. Reach envelope, tool, and a fence line somebody had to argue about.12Infeed side. Everything the robot does is downstream of product arriving correctly.13Palletizing. Vacuum head, slip sheet, cartons squared to the pallet edge.14Layer going down. Pattern geometry, executed a few thousand times a shift without complaint.
The controls behind them
15Cell panel. Everything the arm needs that isn't in the robot controller.16Interior, blue duct, drives and distribution. Built for the technician who opens it at 2am.17I/O landed and labeled. Field wiring in, logic out.18Terminal detail. Pluggable connectors so a card swap isn't a re-wire.
What actually decides a cell
Decision
Why it dominates
End effector
More cells are limited by what the gripper can reliably hold than by anything about the arm. Product variation lives here.
Reach and payload
Chosen at quote, regretted at commissioning. Payload includes the tool, which is routinely forgotten.
Safety architecture
Fencing, light curtains, scanners or collaborative-rated. Changes cost, cycle time and floor space all at once.
Cycle time
Not the arm's spec sheet. The arm waiting on the conveyor, waiting on the sensor, waiting on the operator.
Pallet pattern
Layer patterns and flips are pure geometry, and they will consume more of your week than the robot programming.
Note
Vendors differ more in their teach pendants and their support than in their arms. The mechanical hardware converged years ago; the software and the phone number did not.
A panorama sweep that assumed the world would hold still. It did not. The algorithm
made exactly one bad assumption and then committed to it, with total confidence,
all the way across the frame.
Which is the whole argument for stating a tolerance: every measurement carries one,
whether or not anybody wrote it down.
