Direct answer: Cobot reach is quoted as a radius measured with the arm fully extended, and almost no assembly motion happens in that pose. The useful question is whether every working point can be served in a usable posture. In the reference cycle the arm picks from an upper tray and reaches down into a part on the conveyor below, so two points at very different heights both have to sit inside the envelope.
Who this is for: This guide is written for assembly and process engineers sizing a collaborative robot for a line-side station, where the arm has to serve more than one working point and share space with people.
Scope: It covers how working points, tool geometry, posture margin and clearance for people decide whether an arm is adequate, and what each judgement still owes in evidence. It does not cover gripper design, controller programming, or the risk assessment method itself.
A radius on paper, a set of poses in the cell
The reach figure on a datasheet describes one thing: the furthest point the wrist can occupy with the arm straight. It is a clean, comparable number, and it corresponds to almost nothing a station actually asks the arm to do. Real motions happen with joints folded, with the tool angled, and with something in the way.
So the number is not wrong, it is simply answering a question nobody has. The question a station has is whether the arm can serve each working point in a posture that is usable: stable, not near a joint limit, with the tool oriented as the task requires and nothing colliding on the way in or out.
That difference is why cobot reach has to be checked point by point rather than compared model by model. Two arms with the same quoted radius can behave very differently at the same station once tool length and approach direction are added, and the difference only shows up when the points are named.
Start by listing every working point
In the reference cycle the arm does something very ordinary and very instructive. It reaches up to a parts tray on an upper feeder shelf, takes a small component, then reaches down into a large cast part carried on a pallet on the conveyor below and places the component into position. Two points, quite different heights, different approach directions.
According to ISO 10218-2:2025, what is assessed is the robot application rather than the robot on its own, and cobot reach behaves the same way: it is a property of this station with this tool, not of the arm in a catalogue. The envelope has to contain both points, and it has to contain them in usable postures, not merely touch them. Sizing against the easier of the two, which is the natural thing to do when only one of them is drawn, is how a station arrives at commissioning and needs its feeder moved.
The same applies to every other point the cycle touches: secondary pick positions, inspection presentations, reject drops, tool changes. Listing them with their heights and approach directions turns a vague reach question into an arithmetic one, and it usually takes less time than the first catalogue comparison would.
Posture margin near the boundary
Near the edge of the working volume the arm still reaches, but the number of usable postures collapses. Joints approach their limits, some configurations pass close to singularities, and motion planning gets slower and less predictable. The visible symptoms are sluggish moves, odd paths, and occasional planning failures that are hard to reproduce.
Leaving margin so that routine motions sit comfortably inside the envelope is a selection decision, and it is cheap at selection time and expensive afterwards. In practice this is one of the most common reasons a station that looked adequate on paper needs its layout revisited during commissioning.
According to ISO 9283:1998, robot performance characteristics are determined under defined test conditions, which is precisely why a published figure describes the robot rather than the station. Tool length and orientation shift the working point away from the wrist, and a longer tool pulls the practical boundary in noticeably. The tool concept therefore belongs in the sizing exercise, not after it.
Decision table: what the station layout justifies
The table maps what a station layout can justify on its own against the evidence a supplier still has to produce.
| Evidence from the application | Selection it justifies | Evidence you still owe |
|---|---|---|
| Two working points at similar height, short approach, light tool | A compact arm sized with margin at both points | Pose check at each point with the real tool fitted, including approach and retract |
| Working points at very different heights, as in the reference cycle | Sizing driven by the least favourable point rather than by an average distance | Reach and posture study covering both points and the transition between them |
| Long or angled end tooling, or a tool change during the cycle | Tool geometry fixed before arm selection, since it moves the practical boundary | Envelope check with each tool configuration, including cable and hose routing |
| People passing or working immediately beside the station | Layout and task-based risk assessment before the arm is fixed | Assessed clearance, speed and force conditions for the actual task and layout |
A station without a fence still owes clearance
Collaborative arms are usually installed without a fence, which is much of their appeal and all of their complication. The reference station sits at the side of a conveyor with a walkway beside it, which means the working volume is shared space rather than enclosed space.
That makes layout and safety the same conversation. Where the arm swings, where the tool travels, where the part is presented and where people walk are one drawing, not four. Getting that drawing right early is what keeps a station usable once people are moving around it under production pressure.
According to ISO/TS 15066:2016, collaborative operation is characterised for the specific application rather than by the robot being labelled collaborative, and that is the practical point. The absence of a fence increases what has to be assessed rather than reducing it, and the assessment is of this task, this tool and this layout.
What the footage proves and what it does not
The reference footage proves a visible process. A collaborative arm picks components from an upper tray and places them into a large cast part on a conveyor pallet below, repeating the cycle, with the two working points clearly at different heights. Those are observable facts about the operation.
It proves nothing about the arm model needed, the cycle time, the insertion forces or the safety conditions that apply. Those depend on your points, your tool and your assessed layout, and they cannot be read from a video. EVST records that boundary in the evidence review so that a selection rests on the right inputs.
What is left afterwards is a short and concrete list for cobot reach: a pose check at every point with the real tool, a posture-margin review near the boundary, and a task-based assessment for the shared space. A supplier who delivers those three has answered the reach question; a supplier who quotes a radius has not.
What to put in the enquiry
State the station: a layout with every working point marked, with heights and approach directions, plus the part and tray dimensions and how material arrives. Add the tool concept with its length, mass and orientation, since those move the practical boundary more than most people expect.
Then state the human side: who comes near the station, how often, from which direction and to do what. Where an alternative is under consideration, say so, because a fenced industrial arm, a collaborative robot range option and a shared cell such as a robot machine tending overview case answer the clearance question in completely different ways, as does a path-driven application like collaborative robot path planning. With those inputs, cobot reach and clearance can be checked on one drawing rather than estimated separately.
Frequently asked questions
Can we compare arms on the quoted reach figure?
Only as a first filter. The quoted radius is measured with the arm straight, which is not how assembly motions happen. Compare candidates against your own working points with the real tool fitted, and the ranking often changes.
How much posture margin should we leave?
Enough that routine motions do not run near joint limits or awkward configurations. There is no universal figure, but if the cycle only works with the arm close to full extension at any point, that is a signal to revisit either the layout or the arm.
Does the tool really change the answer?
Yes. The working point sits at the tool tip, not at the wrist, so tool length and angle shift every point outward and pull the practical boundary in. A tool concept fixed after arm selection is one of the more common causes of late layout changes.
Does a collaborative arm remove the need for a risk assessment?
No. It changes what the assessment covers. Without a fence, the shared space, the task, the tool and the layout all enter the assessment, so an unfenced station generally has more to assess rather than less.
Project inputs for an application review
Send the following and reach can be checked at your points instead of compared on a radius:
- a station layout with every working point marked, with heights and approach directions
- part, tray and pallet dimensions, and how material arrives at the station
- the end tooling concept, with length, mass and orientation
- the cycle sequence, including any tool change or inspection presentation
- who approaches the station, how often, and from which direction
If you are sizing a collaborative arm for a line-side station, send the layout with all working points and their heights, the part and tray dimensions, the tooling concept, the cycle sequence, and the movement pattern of people around the station. Reach, posture margin and clearance can then be checked on one drawing. Related reading: collaborative robot path planning, robot machine tending overview.