Cobot payload and reach: how to size a cobot.

Spec sheets print one number because one number fits in a table. A UR20 is "20 kg". A GoFa 5 is "5 kg". A CRX-10iA is "10 kg". Each of those is true, and each comes with conditions that live a page or two later in the datasheet or product specification.

The first condition is what counts. The payload is the whole stack on the tool flange. That's the gripper, the quick changer or adapter plate, any camera or force sensor riding on the wrist, the part, and the short run of cable or air hose that hangs off the flange before the first clamp. Universal Robots' own guidance is to enter both the tool and the work item when you set payload mass and center of gravity. If your gripper weighs 1.1 kg, a 5 kg arm is a 3.9 kg arm for parts.

The second condition is where the load sits. Universal Robots says it plainly in a 2026 blog post titled "Why 20 kg is never really 20 kg": spec sheets assume the load sits near the flange center, and real grippers and tool changers often push the center of gravity 200 to 300 mm out. ABB's GoFa specification draws the rated mass inside a small box of vertical (Z) and radial (L) offsets and notes that the diagram itself assumes a nominal load inertia.

The third condition is pose and dynamics. Repeatability numbers are measured per ISO 9283, the standard test method for manipulator performance, at a defined load. ABB, for example, states its GoFa path accuracy "at rated maximum load, maximum offset and 1.6 m/s". Change the load and the performance numbers move with it.

Picture the forearm and wrist as a cantilever. The load hangs off the end, and the motors in joints 4, 5 and 6 have to hold it against gravity and then accelerate it. The torque they need is the weight times the moment arm. Double the distance and you double the static torque, before you've moved anything.

Vendors handle this in two ways. Universal Robots prints a payload curve in every datasheet: full rated payload flat out to some offset, then a falling line. FANUC publishes allowable moment (in N·m) and allowable inertia (in kg·m²) for the J4, J5 and J6 axes on its CRX pages. ABB draws a two-dimensional load diagram for GoFa, with mass contours plotted against Z and L offsets, plus inertia formulas for axes 5 and 6. Different formats, same physics.

One honest caveat about the charts. They're printed as drawings, so any value you read between gridlines is approximate. When the operating point is near the line, don't argue about the third digit. Pick the bigger arm or move the load closer.

Mass decides how much torque it takes to hold the load still. Inertia decides how much it takes to rotate it quickly. The relationship is the rotational version of F = m·a. Torque equals inertia times angular acceleration, so for a fixed motor torque, double the inertia and you halve how fast the wrist can speed up and slow down.

Two loads can weigh the same and behave very differently. A 4 kg steel puck held at the flange is easy. A 4 kg aluminum sheet 600 mm wide is a sail, and spinning it around joint 6 takes far more torque, because inertia grows with the square of distance. Rotate the sheet by 90 degrees on every cycle and you'll find the limit fast.

Not every vendor publishes an inertia limit, which doesn't mean there isn't one. ABB gives formulas for GoFa. For axis 5 the load's inertia is its mass times the squared distance from the axis (ABB's formula adds fixed offsets of 0.101 m and 0.08 m to Z and L for the wrist geometry) plus the load's own inertia about its center. The limit is 0.35 kg·m² on the GoFa 5 and 0.58 kg·m² on the 10 and 12. FANUC lists allowable inertia per wrist axis on the CRX pages, for example 1.28, 0.9 and 0.3 kg·m² for J4, J5 and J6 on the CRX-10iA. Universal Robots' datasheets don't print a separate inertia limit, so for wide or long tools, ask UR or your integrator for a check with your CAD numbers.

Start with what the reach number measures, because it isn't the same everywhere. ABB quotes GoFa in two ways: 0.95 m to the wrist center and 1.05 m to the flange on the GoFa 5. Universal Robots and FANUC give one reach figure per model. Whichever it is, it's measured to the robot, not to your fingertips, and it describes a point the robot reaches with its arm stretched out.

The task point is somewhere else. If the gripper is 180 mm long and must approach straight down, the flange has to be 180 mm above the part, so a part on a low table near the edge of the sphere is out of reach even though its distance from the base is less than R. Add the approach and retract moves, which need room above and in front of the part, and the usable volume shrinks again.

Then there's speed. Near the edge of the envelope, small moves of the tool need large moves of the joints, and near the column above the base the wrist joints can be asked to spin very fast. Both cost cycle time. We put the machine door, the infeed and the outfeed inside the ring in Fig. 03, and we treat a layout that needs more than about 85 percent of rated reach as a reason to pick a longer arm, add a riser or move the robot. Our machine tending cells are laid out this way before anyone picks the arm.

The job is loading an aluminum housing into a vise. The heaviest variant is 4.0 kg. It's gripped by a 1.1 kg electric two-finger gripper with long fingers, on a 0.3 kg tool changer. From CAD, the changer's center of gravity is 12 mm from the flange, the gripper's is 55 mm, and the part's is 200 mm down the tool axis and 81 mm to the side, because the fingers grip it near one end. The numbers are illustrative, but they're typical of the tending jobs we scope.

The pick point on the conveyor is 780 mm from the robot base axis, and the vise is 690 mm out. We'll check two candidates first, a Universal Robots UR7e (7.5 kg, 850 mm) and an ABB GoFa 10 (rated 10 kg, 1.52 m to the wrist center), then see where they land.

Mass is easy, and it's where most people stop. 5.4 kg against 7.5 kg looks like plenty of headroom. The center of gravity is where it gets interesting. The part dominates the average, so the combined center lands 160 mm out, past the 150 mm where the UR7e's flat line ends. The curve still allows about 7.4 kg there, so the UR7e passes on payload with about 27 percent margin.

Then reach kills it. 780 mm is 92 percent of 850 mm, with a 200 mm long tool that needs to come down onto the conveyor. That's the edge of the sphere, and it'll be slow and awkward if it works at all. A UR12e (12.5 kg, 1,300 mm) puts the pick at 60 percent of reach and carries about 11.7 kg at this offset. The GoFa 10 passes on reach too, but the inertia check comes out at 92 percent of ABB's limit, so we'd want ABB to confirm it before committing.

A few things in that table are worth slowing down for. The UR30's flat line is the shortest on the list, which makes sense for a compact, heavy-payload arm built for things like screwdriving and close-in palletizing. The UR20 trades some payload for a longer flat line and much more reach. And ABB's "wrist-down" ratings, 12 kg on the GoFa 10 and 14 kg on the GoFa 12, apply only with the wrist pointing straight down and the load within a few centimeters of the flange, which fits a vacuum pick but not much else.

FANUC's moment limits translate directly. The CRX-10iA's J6 limit of 11 N·m means a 10 kg load can sit no more than about 112 mm off the J6 axis (11 ÷ (10 × 9.81)), before you count anything about the J5 geometry. Always use the vendor's own load diagram or sizing tool for the final answer; the arithmetic here is for screening.

If you're comparing brands on these numbers, our head-to-heads go deeper on Universal Robots vs FANUC CRX, Universal Robots vs ABB GoFa and Universal Robots vs Doosan.

Sizing from the model name. "It's a 10 kg robot and the part is 6 kg" is how most undersized cells start. Run the center-of-gravity math with the real gripper before the quote is signed.

Weighing the average part. Size for the heaviest variant you'll ever run in that cell, plus whatever someone will bolt to the wrist next year. A camera bracket and a second sensor add up.

Leaving the payload setting at the default. Universal Robots says the active payload mass and center of gravity are used to control the joint motors, and that incorrect settings can cause incorrect movements and protective stops. ABB goes further: wrong load data can overload motors, gearboxes and structure, running outside the load diagram isn't covered by warranty, and its power and force limiting functions need the payload correctly defined. Update the payload when the robot picks and drops parts too, since the load changes mid-cycle. Both vendors offer an automatic identification routine; use it as a check on the CAD numbers.

Treating reach as a promise. Simulate every pose with the real tool and approach moves. If the layout needs the last 15 percent of the sphere, move the robot or buy reach.

Forgetting the safety side. In power and force limited operation, a heavier load carries more energy into a contact, and that can force lower speeds than the cycle time assumed. Our guide to cobot safety walks through how that's assessed, and the ISO 10218 and R15.06-2025 changes matter here too.