Collaborative Robots
Why Cobots Run Slower and What That Costs You
Collaborative operation caps tool speed by body region and effective mass. Here is the arithmetic, the throughput you give away, and when a fence is simply cheaper.
A collaborative cell is slower because the permissible tool speed is set by the energy a human body region can absorb, not by the motors. The published worked example allows 2,400 mm/s toward a hand at 1 kg effective mass but only 400 mm/s toward a chest at 20 kg, while a fenced arm of the same size routinely runs at 2,000 mm/s to 3,000 mm/s throughout the cycle.
The mechanism, in one paragraph
In contact-permitted operation the robot is allowed to strike a person, so the design constraint becomes the energy transferred in that strike. The standard models the collision as a fully inelastic two-body contact: robot effective mass against body-region effective mass, through a spring constant that represents the tissue. Set the transferable energy to the published limit for that body region, solve for relative velocity, and the answer is your speed cap. Nothing about motor power enters the calculation.
Three variables move the answer, and only one of them is under the programmer's control.
- Body region. The energy limit ranges from 0.11 J at the face to 2.6 J at the pelvis, a factor of 24.
- Robot effective mass. A function of posture, moving links and payload. Higher mass, lower permissible speed.
- Contact area. The published table assumes 1 cm². A sharper edge concentrates pressure and pushes the limit down further.
What the cap actually is
| Exposed body region | 2 kg arm | 5 kg arm | 10 kg arm | 20 kg arm |
|---|---|---|---|---|
| Hand and finger | 2,200 | 2,000 | 2,000 | 1,900 |
| Lower arm | 1,800 | 1,500 | 1,400 | 1,300 |
| Abdomen | 2,100 | 1,400 | 1,000 | 780 |
| Upper leg | 1,400 | 920 | 670 | 500 |
| Shoulders | 1,200 | 790 | 590 | 450 |
| Chest | 1,100 | 700 | 520 | 400 |
Read the columns, not the rows. The penalty for mass is brutal in the torso rows: an abdomen tolerates 2,100 mm/s from a 2 kg arm and 780 mm/s from a 20 kg arm, a drop of 63 %. The hand row barely moves, falling only 14 % across the same range, because a hand has an effective mass of 0.6 kg and simply gets pushed away.
Translating the cap into parts per hour
A cycle is not one long move. It is a sequence of approach, fine positioning, process time and retract, and the speed cap only touches the free-space segments. That is why the throughput loss is usually smaller than the speed ratio suggests, and why measuring it properly changes procurement decisions.
| Segment | Fenced, 2,500 mm/s | Collaborative, 700 mm/s |
|---|---|---|
| Approach to bin, 1,100 mm | 0.9 s | 2.4 s |
| Fine positioning and grasp | 1.4 s | 1.4 s |
| Transfer to chuck, 1,400 mm | 1.1 s | 2.9 s |
| Insert, clamp, release | 2.6 s | 2.6 s |
| Retract, 900 mm | 0.8 s | 2.0 s |
| Cycle total | 6.8 s | 11.3 s |
| Parts per hour | 529 | 318 |
The speed ratio is 3.6 to 1. The throughput ratio is 1.66 to 1, because 4.0 s of the cycle is process time that does not care about guarding. Over a 250-day year on two shifts that difference is roughly 844,000 parts against 507,000 parts, and whether that gap matters depends entirely on demand.
How to recover most of the loss
Cells rarely have to run at the worst-case cap for the whole cycle. Four techniques recover a large share of the gap without changing the safety case.
- Zone the speed. Run at full speed where no body region can be reached, and drop to the cap only inside the collaborative workspace. Many cells spend 60 % to 80 % of their path outside it.
- Raise the exposed region. A guard plate that makes the torso unreachable moves the binding row from chest to hand and can triple the permissible speed.
- Increase contact area. Padding and rounded covers lift the pressure limit, which raises the derived speed for the same energy budget.
- Switch operating type mid-cycle. Speed and separation monitoring during transfer, contact-permitted only during the final approach, is a common and legitimate pattern.
The one thing that does not work is asking the vendor for a faster arm. The cap is a property of the collision, and a stronger motor changes nothing on the left side of the equation.
Frequently asked questions
How much slower is a cobot than a fenced robot?
On the free-space segments, typically three to six times slower depending on the exposed body region and effective mass. On the full cycle, usually 1.4 to 2 times slower, because process time and fine positioning are unaffected by the speed cap.
Is the 250 mm/s limit the collaborative speed limit?
No. The 250 mm/s figure in ISO 10218-1 applies to manual reduced-speed mode, used while teaching. Collaborative operating speeds come from the contact model and are usually much higher, from roughly 400 mm/s to 2,900 mm/s.
Can I speed the cell up by reducing payload?
Partly. Lower payload reduces effective mass, which raises permissible speed, and the effect is strong in the torso rows. Going from 20 kg to 5 kg effective mass raises the chest limit from 400 mm/s to 700 mm/s, a 75 % gain, while the hand limit barely moves.
Does speed and separation monitoring avoid the penalty?
It shifts it. The robot may run fast while the gap is large, but the protective separation distance grows with speed and with detection latency, so a fast arm needs a large monitored area. In cramped cells the resulting stop distance often forces the same slowdown by a different route.
When is a fence simply the right answer?
When the cell is throughput-bound, when the tool is the hazard rather than the arm, when payload exceeds roughly 35 kg, or when the product runs unchanged for years. Guarding is a one-time cost that buys back the full cycle speed.
Sources
- ISO/TS 15066:2016, Annex A, Tables A.4 and A.5Energy limits and the transient contact speed table quoted throughout
- ISO 10218-1:2025, Robotics, Safety requirements, Part 1International Organization for Standardization, reduced speed and collaborative operation
- Collaborative robot systems, information sheet 080German Social Accident Insurance (DGUV), zoning and validation practice