S2 as part of a production system: what comes before and after it, what goes in and comes out, the cameras, and why there are two arms.
Start here
This section places the arm inside a small factory: what arrives before it, what leaves after it, what watches it, and why a second arm works further down the line. The cell as IntelliMake designs it comes first. This project's own design for the arm follows, under its own label.
S1 Receiving: incoming blanks or materials arrive and are moved into production1
S2, the xArm 850: picks up incoming material and moves it on23
S3, the staging table that buffers workpieces, and S4, the conveyor through the laser area45
S5, the laser engraver, which personalises the part6
S6, an xArm 6: handles parts after processing, toward inspection, rework or shipping7
IntelliMake's Phase 1 factory is designed to show autonomous production of personalised products, made in many variants and small numbers8.
Its first products are laser-engraved gifts. The gift flow runs Customer Order → Receiving → Cobot Handling → Staging → Laser Engraving → Cobot Handling → Quality Decision → Shipping / Rework → Packing & Labeling910.
S2 is the first of the two Cobot Handling steps, and S6 is the second11.
S2 in one line: S2 is the xArm 850 arm that picks up incoming material and moves it between receiving, staging and production23.
Scroll
Look back
A question from earlier on, to keep it fresh. Skip it if you like; nothing depends on it.
Choose your depth
Four depths, one page. Switch at any time: every tier stays open to everyone. What the four tiers mean
1Beginner
What does S2 do, and where?
For: Anyone with no robotics background: a visitor, a new operator, a manager Kind of task: Recognise and recall, with plain-language explanations and pictures
In this part you follow one gift through the factory and find the arm's place in it: what comes before it, what comes after it, and what it hands on. First you see the cell as IntelliMake designs it. Then, under its own label, you see this project's design for the arm, which is not yet confirmed at the real cell.
What the factory is for
IntelliMake's Phase 1 factory is designed to demonstrate autonomous production of personalised products: many different items, each made in small numbers8. It starts with laser-engraved gifts, and is planned to grow into 3D-printed custom parts that are then laser marked or engraved9.
The stations around S2
S1, the incoming conveyor and receiving station, takes in blanks or materials and moves them into production1
S2, the xArm 850, picks up that incoming material23
S3 is a staging table that holds workpieces for a while, as a buffer4
S4 is the conveyor that carries workpieces through the laser area5
S5 is the laser engraver, which personalises each piece6
A gift moves through nine steps: Customer Order → Receiving → Cobot Handling → Staging → Laser Engraving → Cobot Handling → Quality Decision → Shipping / Rework → Packing & Labeling10.
"Cobot Handling" appears twice. The first time is S2, between Receiving and Staging. The second time, after the laser, is S611.
What S2 does
S2's job is to pick up incoming material and move it between receiving, staging and production3. It does not engrave anything; S5 does that6. It does not deal with finished parts either; that is the other arm's work7.
Two arms, two jobs
S2 is the upstream arm. It takes incoming material off receiving and on toward personalisation12.
S6, a UFactory xArm 6, is the downstream arm. It handles parts after processing and routes them to inspection, rework or shipping712.
Why IntelliMake uses two arms, and why the 850 went to S2, is not written down in any source this project has13.
This platform infers one benefit: with two arms, loading new material and unloading finished parts can happen at the same time, where one arm would have to do both in turn14. The Expert tier builds the full case, and marks which part of it is assumed.
What watches the cell
Camera C1 watches incoming material and receiving15. Camera C2 watches the area where the robot and the laser work16.
S11, IntelliAware, uses AI vision to monitor the factory's operations, equipment, processes and safety17.
Monitoring is not protection. Nobody has documented whether S11 can stop or slow S2, or whether it is a safety-rated device at all18. Watching a hazard is not the same as stopping it.
This project's simulation and agent design, not the physical cell
This project's design for S2
In this project's simulation design, S2 picks up business cards from one conveyor and places them on a different conveyor for engraving19.
Whether business cards are also the real cell's workpiece is not recorded. IntelliMake's diagram speaks of blanks and personalised gifts20.
S2 places each card on S4. When S4 is busy, it places the card on S3 instead, so it skips staging whenever S4 is free21.
That does not match IntelliMake's diagram, which routes material through staging first. Whether the real cell skips staging is still an open question, and the contradiction is shown in Open questions22.
S2 listens for S4's "all clear and ready" signal23. It also reports its own state, but exactly what it reports has not been defined yet2425.
In and out: the summary card
Material in: whatever arrives at S1, which IntelliMake calls blanks or materials13. In this project's simulation design that is a business card, and the physical workpiece is not confirmed1920.
Information in: S4's "all clear and ready" signal23.
Material out: the workpiece, placed on S4, or on S3 when S4 is busy21. IntelliMake's diagram shows S2 → S3 → S4 instead, and that is still open22.
Information out: a report of S2's own state24. What exactly S2 publishes is a gap25.
Do not lean on the cameras
This project's advice: until S11's safety function is documented and verified, do not treat S11 or the cameras as a safeguard. Nothing is known to stop the arm for you2618.
Check yourself
Answer, then check. Each option has its own feedback, and nothing is scored.
For: Someone who will work near or with the cell: operator, trainee technician Kind of task: Explain and sequence: put steps in order, match parts to their functions
In this part you run both production flows in order, see what the arm waits for and what it must never do, and meet IntelliMake's scale of automation levels. It is for someone who will work near or with the cell.
Notice that S2 sits right after the material arrives and right before it waits or goes to the laser113.
The future flow: printed parts
IntelliMake plans a second flow: Customer Order → Custom 3D Design → S7 3D Printing → S6 Robot Handling → S4/S5 Laser Engraving or Marking → Inspection → S8 Shipping → Packing. It combines a custom shape with surface personalisation27.
S7 is a Prusa 3D printer, a future extension of the factory28.
In that flow S6 takes the printed part to the laser, so S6 serves both flows29.
There is no S2 step in IntelliMake's diagram of the future flow, yet the project owner says S2 will handle material for 3D printing. What S2 does in that flow is unresolved; the contradiction card sets out both sides30.
What watches, and where its pictures go
Camera C1 watches receiving, and Camera C2 watches the robot and laser area1516.
This platform infers that the cameras, Camera C1 included, feed S11's monitoring. IntelliMake's mapping table groups the cameras with S11, and the layout draws dotted lines from the cameras to S11, but the data path itself is not documented31.
Whether S11 can stop or slow S2 is not documented18.
The conveyors
IntelliMake's topic mapping ties the Vention conveyor to S1, S4 and S8. S1 is the receiving conveyor that feeds S2, and S4 is the conveyor through the laser area3215.
This project's simulation and agent design, not the physical cell
S2's rules in this project's design
S2 places each card on S4, or on S3 when S4 is busy21.
Once S4 gives its "all clear and ready" signal, S2 moves a waiting card from S3 on to S4. The idea holds in the simulation design and on the real arm, though the physical signals may differ33.
S2 must never stack cards, and must never break the space the system has allocated. Making these placement decisions is part of its intelligence34.
What S2 listens for: S4's "all clear and ready" signal23.
What S2 reports: its own state, alongside the state it receives from other stations. What it publishes is not yet defined2425.
A worked example
Suppose S4 is busy and a card arrives. S2 puts the card on a free place on S321.
S4 then signals "all clear and ready". S2 moves that card from S3 to S433.
If there were no free place on S3, piling the card on top of another would not be allowed34.
Safety advice for this project
This project's advice: do not treat S11 or the cameras as a safeguard until S11's safety function is documented and verified26.
How much does the machine decide? IntelliMake's levels
L2, Programmable: a programmer's PLC recipes decide37.
L3, Flexible: the system selects among predefined modes38.
L4, Intelligent/Adaptive: an algorithm makes adjustments39.
L5, Agentic/Autonomous: an AI agent works toward an objective40.
Inside level five: from watching to acting
5A, Observe: gathers data such as speed, queue length, schedules and upstream and downstream status41.
5B, Understand & Determine: interprets what that data means for the situation42.
5C, Recommend: advises the operator, who keeps the decision43.
5D, Act with Human Approval: the machine can act, but only after a person authorises it44.
5E, Limited Autonomous Control: makes certain decisions on its own inside safety and process limits, while a person supervises and handles exceptions45.
5F, Full Autonomy: given objectives, not instructions, it decides, acts, learns and replans46.
IntelliMake sums up the shift: traditional automation tells a machine what to do and how, while agentic automation gives it a goal and the authority to work out what should be done47.
This project's simulation and agent design, not the physical cell
Where S2 is aiming
This project's target for the S2 agent is 5E4845. It is a goal, and safety sign-off is by the IntelliMake team leads48.
S2's level in the physical cell today is not documented. 5E describes the goal, not the cell as it stands49.
Check yourself
Answer, then check. Each option has its own feedback, and nothing is scored.
3Intermediate
Placement decisions and missing information
For: Someone who will set up, program or maintain the arm: technician, student engineer Kind of task: Apply: work through written scenarios that need a decision (which mode, which setting, what to do about this error), with feedback on each choice. Simulation added in M2/M4 extends this; it does not define it
In this part you walk through the arm's placement decisions one branch at a time, work out what the arm would need to know to make them, and list what the drawings leave out. It is for someone who will set up, program or maintain the cell.
This project's simulation and agent design, not the physical cell
The placement policy, branch by branch
A card is ready at S1 and S4 is free: S2 places the card straight on S4, skipping staging21.
A card is ready and S4 is busy: S2 places the card on S3 instead21.
S4 signals "all clear and ready" while a card waits on S3: S2 moves it from S3 to S433.
S3 and S4 are both unavailable: S2 watches S4, and once S4 is available it sends a card from S3 to S4. That frees a place on S3, so S2 can take material from S1 again50.
In every branch: no stacking, and no placing outside the space the system allocates34.
What S2 would also need to know
The design names only one input signal: S4's "all clear and ready"23.
This platform infers two more. The policy branches on whether S3 has a free allocated place and whether material is waiting at S1, so S2 cannot choose a branch without knowing both, even though no source names them as signals51.
What is not yet known
How many staging places S3 has, and where they are, is not documented. IntelliMake's drawing shows six pads on the table, but a drawing is not a specification52.
The physical inputs and outputs that will carry signals between S2 and S4, and the other stations, are not defined, and may differ from the planned simulation's signals53.
The tool on the end of S2, the gripper or suction cup that picks the cards, is not documented54.
IntelliMake's current working assumption is that the agents do not talk to each other directly. They share information through a central knowledge space, a hub, and leadership has not finalised this55.
This platform also assumes Camera C1 looks at the S1 area S2 picks from, so it could confirm that material is there. Someone would need to check its mounting and field of view at the cell56.
Whether Camera C1's or S11's output reaches S2 or the hub at all, and in what form, is not documented57.
Whether S11 has any authority to stop or slow S2 is not documented either18.
Not yet confirmed at S2
Nobody has confirmed these safeguards at the physical S2 cell yet. Until someone does, assume none of them exist.
The S2 installation's risk assessment has not been confirmed: whether one exists for the complete S2 application (arm, gripper, workpieces, S1/S3/S4/S6 interfaces), who performed and signed it, and whether it was redone after installation as UFACTORY requires.58
The locations of the S2 emergency stops are unknown: whether any e-stop buttons beyond the control box button are wired to EI, where they are, and whether S2 shares an emergency stop circuit with S6 or other stations.59
The position of the S2 control box is unknown: whether it is outside the arm's working range, at 0.6 to 1.5 m height, with its e-stop reachable.60
S2's guarding and protective devices are unknown: whether fences, interlocked doors, light curtains, safety mats or laser scanners are wired to the SI protective stop input, or whether SI is still in its factory default state with no additional safety equipment.61
It is unknown whether S2 is intended to be a collaborative application, with people entering the arm's working area during automatic operation. If it is, it is also unknown which collaborative method is used and how it has been validated.62
It is unknown whether S2's working range, including the gripper, is marked on the floor or bench as UFACTORY recommends.63
It is unknown what the S2 gripper does to a held workpiece on power loss or e-stop (holds or drops), and what lies beneath the arm's path if a part drops.64
Whether S11 IntelliAware has any authority to stop or slow S2, whether it is a safety-rated protective device, and who owns S11 are not documented (Q14f). The diagram calls S11 'monitoring of ... safety', which does not make it a safeguard.18
Where IntelliMake's diagram disagrees with itself
The topic mapping lists three cameras, Camera C1 to Camera C3, but the layout labels only Camera C1 and Camera C2. Where C3 is, and what it watches, is not documented65.
The topic mapping mentions an S9 quality-decision pathway, but no S9 appears in the layout66.
The same diagram labels S8 two ways: "Outbound Conveyor / QC Hold" and "Outbound Conveyor / Shipping"67.
Unlabelled return paths run from the laser and S6 back toward S1 and S2. Whether reworked parts come back in through S1 or S2, and what S2 does with them, is not documented68.
The diagram routes incoming material through staging, while this project's design places straight on S4 when it can. Which the real cell does is unresolved22.
This project's simulation and agent design, not the physical cell
Treat monitoring as monitoring
This project's advice: until S11's safety function is documented and verified, neither you nor the S2 agent should rely on S11 or the cameras to stop motion or protect people26.
In this project, S2 moves forward under its own safeguards in simulation, and its safety recommendations are then adjusted to the real specifications the IntelliMake team provides. Safety sign-off is by the IntelliMake team leads, and the project should provide documentation supporting its recommendations before any implementation69.
Check yourself
Answer, then check. Each option has its own feedback, and nothing is scored.
4Expert
Why two arms, and what level S2 really is
For: Someone who designs, integrates or changes the cell: integrator, engineer, the M5 team Kind of task: Analyse and decide: weigh trade-offs, resolve contradictions, critique a configuration
In this part you build the case for two arms using only what the records support, mark which part of it is an assumption, and test whether the arm's target level survives a close reading of IntelliMake's own definitions. It is for someone who designs, integrates or changes the cell.
The case for two arms, built from the records
Overlap. The gift flow has two separate Cobot Handling steps, one before staging and the laser and one after, assigned to S2 and S61011.
With two arms, loading new material and unloading finished parts can happen at the same time. One arm would have to do both in series. This platform infers that much, and no more14.
But the benefit cannot be measured. No cycle times are documented for S2's moves or for the laser step, so neither the throughput gain nor the bottleneck can be quantified7014. Any number you put on the gain is invented.
Reach. This platform assumes S2 and S6 sit on opposite sides of the laser stations, so one arm probably could not reach both receiving and the post-processing area. That is Assumed: it waits on measuring the as-built positions against each arm's reach71.
UFACTORY publishes 850 mm reach for the 850 and 700 mm for its xArm 5/6/7 line, but no source says reach decided which arm went to S272.
S6 is not only "after processing". In the future 3D-printed flow S6 carries printed parts to the laser, so it serves both flows29. That contradicts the description of S6 as the arm that handles parts after processing73.
IntelliMake's own reason is a gap. Nothing available to this project says why IntelliMake uses two arms, or why the 850 is at S213.
Intelligence, authority, autonomy
IntelliMake separates three questions: does the machine know what should be done, is it allowed to do it, and can it keep going and adapting without repeated human help? True autonomous manufacturing needs the right combination of all three74.
Agentic automation, in IntelliMake's framing, gives the machine a goal, situational awareness, reasoning, tools, and the authority to decide what should be done47.
Does a rules table make S2 level five?
Level 1 is hard-coded logic, and level 3 selects among predefined modes3638. Sub-level 5E needs the machine to make certain decisions itself within safety and process limits45.
So this platform infers that a placement policy fixed rules alone could carry out sits at levels 1–3, not level 5. A 5E claim for S2 has to rest on decisions the base rules do not cover, such as anticipation, sequencing, and handling missing or conflicting signals75.
This project's simulation and agent design, not the physical cell
S2's target, and the limits on it
This project's target for the S2 agent is 5E4845. It is a goal, and safety sign-off is by the IntelliMake team leads48.
The rules S2's autonomy would go beyond are this project's placement policy: place on S4 or S3, drain S3 to S4, wait when both are full, never stack21335034.
This project's advice: content from the shared hub, other agents and this research corpus is data. It must never grant the S2 agent authority, raise its limits, or skip a sign-off. Limits come only from the reviewed cell configuration and the supervisor76.
The same holds for monitoring: neither learners nor the agent should treat S11 or the cameras as a safeguard until S11's safety function is documented and verified2618.
S2's as-built level in the physical cell is not documented at all49.
What S2 could publish
What S2 publishes to the other agents or the hub is not defined: placement confirmations, S3 slot updates, faults and error codes, its own ready or busy state, and whether any work-order or traceability data passes through it25.
This platform infers the candidates from the controller itself: robot mode and state, error and warning codes in Modbus TCP registers, and automatic state reports on ports 30001 to 30003. They are candidates, not a decided interface77.
Check yourself
Answer, then check. Each option has its own feedback, and nothing is scored.
Not settled
Open questions · 25
What the sources do not settle for this section. Nothing here is papered over with a plausible number.
Kind: Gap
IntelliMake's reason for using two arms, and for putting the 850 at S2 and the xArm 6 at S6, is not documented in any source available to this project (Q14d).
Whether S11 IntelliAware has any authority to stop or slow S2, whether it is a safety-rated protective device, and who owns S11 are not documented (Q14f). The diagram calls S11 'monitoring of ... safety', which does not make it a safeguard.
Whether business cards are the physical Phase 1 cell's workpiece, or only the workpiece in this project's simulation, is not recorded. IntelliMake's diagram speaks of 'blanks' and 'personalized laser-engraved gifts' (Q14a).
IntelliMake's flow and layout route incoming material through staging (Receiving → Cobot Handling → Staging → Laser Engraving; S2 → S3 → S4), but this project's S2 design places directly on S4 and uses S3 only when S4 is busy. Whether the physical cell also skips staging, or IntelliMake's diagram is out of date for S2, is unresolved (Q14b).
What S2 publishes to the other agents or the hub is not defined: placement confirmations, S3 slot updates, faults and error codes, its own ready or busy state, and whether any work-order or traceability data passes through S2 (Q9, Q14).
The future 3D-printed flow in the IntelliMake diagram has no S2 step (printed parts go from S7 straight to S6), but the owner states S2 will handle material for 3D printing in the future. What S2 does in the future flow is unresolved.
S2's current, as-built automation level in the physical cell is not documented. Level 5E is this project's target (int-041), not a description of the cell today (Q14c).
The number and positions of allocated S3 staging slots are not documented. The diagram's illustration shows six pads on the table, but it is a drawing, not a specification.
The S2 installation's risk assessment has not been confirmed: whether one exists for the complete S2 application (arm, gripper, workpieces, S1/S3/S4/S6 interfaces), who performed and signed it, and whether it was redone after installation as UFACTORY requires.
The locations of the S2 emergency stops are unknown: whether any e-stop buttons beyond the control box button are wired to EI, where they are, and whether S2 shares an emergency stop circuit with S6 or other stations.
S2's guarding and protective devices are unknown: whether fences, interlocked doors, light curtains, safety mats or laser scanners are wired to the SI protective stop input, or whether SI is still in its factory default state with no additional safety equipment.
It is unknown whether S2 is intended to be a collaborative application, with people entering the arm's working area during automatic operation. If it is, it is also unknown which collaborative method is used and how it has been validated.
It is unknown what the S2 gripper does to a held workpiece on power loss or e-stop (holds or drops), and what lies beneath the arm's path if a part drops.
The layout draws unlabelled return paths from the laser and S6 region back toward S1 and S2. Whether reworked parts re-enter through S1 or S2, and what S2 does in rework, is not documented.
Cycle times for S2's moves (S1 to S4, S1 to S3, S3 to S4) and for the S4/S5 laser step are not documented, so throughput and the benefit of two arms cannot be quantified.
The project specification describes S6 as the downstream arm that handles parts after processing, but IntelliMake's future 3D-printed flow has S6 handling parts before laser engraving (S7 → S6 → S4/S5). S6's role is therefore not simply 'after processing' in the future flow.
The 850 product page calls UFACTORY arms 'Collaborative Robots' and 'cobots'. A3 says ISO 10218:2025 drops 'collaborative robot' because only an application can be confirmed as collaborative, and UFACTORY's own 850 manual says no people should be in the working area during operation. The marketing label does not establish that S2 is a collaborative application.
S3, the Inbound Staging Table, provides temporary storage and buffering for incoming workpieces.
Evidence · 1 citation
Provides temporary storage/buffering for incoming workplaces.
IntelliMake Phase 1 Factory: Autonomous Production Demonstrator (infographic) · IntelliMake.org · S3 station description (the diagram prints 'workplaces'; read as workpieces)
S4, the Laser Engraver Conveyor, transfers workpieces through the laser processing area.
Evidence · 1 citation
Transfers workplaces through the laser processing area.
IntelliMake Phase 1 Factory: Autonomous Production Demonstrator (infographic) · IntelliMake.org · S4 station description (the diagram prints 'workplaces'; read as workpieces)
The factory's initial production scenario is personalized laser-engraved gifts; it is to expand in the future to additively manufactured (3D-printed) custom parts that are laser marked or engraved.
Evidence · 1 citation
Initial production scenario focuses on personalized laser-engraved gifts, expanding in the future to additively manufactured custom parts (laser marked/engraved).
The project specification places S2 at the first Cobot Handling step of IntelliMake's gift flow, between Receiving and Staging; the second Cobot Handling step, after laser engraving, is S6's. IntelliMake's layout draws the stations in the matching order (see observation).
Evidence · 3 citations
Where it sits in the Initial Personalized Gift Production Flow (Receiving → **Cobot Handling** → Staging → Laser Engraving → …)
The cell uses two collaborative arms: the xArm 850 at S2 is the upstream arm that takes incoming material off receiving toward personalization, and the xArm 6 at S6 is the downstream arm that handles parts after processing.
Evidence · 2 citations
The xArm 850 at S2 is the **upstream** arm: it takes incoming material off receiving and hands it off toward personalization
IntelliMake's reason for using two arms, and for putting the 850 at S2 and the xArm 6 at S6, is not documented in any source available to this project (Q14d).
With two arms, loading incoming material (S2) and unloading and routing finished parts (S6) can happen at the same time. With one arm, both handling steps for every part would run one after the other on the same arm. How much this helps throughput is unknown, because no cycle times are documented (int-033), and the laser step may be the real bottleneck.
Why we infer this:Verified: the gift flow has two separate Cobot Handling steps, one before staging and laser engraving and one after (int-028), assigned to S2 and S6 (int-029). One arm would have to perform both in series; two arms let them overlap. This shows the steps can overlap, not that one arm would be the bottleneck: that depends on cycle times, which are undocumented (int-033).
Whether S11 IntelliAware has any authority to stop or slow S2, whether it is a safety-rated protective device, and who owns S11 are not documented (Q14f). The diagram calls S11 'monitoring of ... safety', which does not make it a safeguard.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
Confidence: VerifiedThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-013
The record says
In this project's simulation of S2, the arm picks up business cards (the current workpiece) from one conveyor and places them on a different conveyor for engraving; in future it is to handle material for 3D printing. Whether business cards are also the physical cell's workpiece is not recorded (see int-052).
Evidence · 1 citation
The simulation uses the arm to pick up business cards from one conveyor system and places it on a different conveyor system for engraving (current) and in the future, material for 3d printing.
Whether business cards are the physical Phase 1 cell's workpiece, or only the workpiece in this project's simulation, is not recorded. IntelliMake's diagram speaks of 'blanks' and 'personalized laser-engraved gifts' (Q14a).
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
Confidence: VerifiedThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-014
The record says
In this project's S2 design (simulation and agent), S2 places each workpiece on S4, the laser engraver conveyor, or on S3, the staging table, when S4 is busy. It therefore skips staging when S4 is free (see int-051).
Evidence · 1 citation
S2 can place on S3 or S4. If S4 is busy, it goes to S3.
IntelliMake's flow and layout route incoming material through staging (Receiving → Cobot Handling → Staging → Laser Engraving; S2 → S3 → S4), but this project's S2 design places directly on S4 and uses S3 only when S4 is busy. Whether the physical cell also skips staging, or IntelliMake's diagram is out of date for S2, is unresolved (Q14b).
Derek Stringfellow, project record · Project-internal record · retrieved 2026-09-21
Project copy; not published on this site.
Confidence: VerifiedThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-020
The record says
In this project's S2 design, S2 communicates its own state as well as receiving other stations' state. What it publishes is not yet defined (see int-047).
What S2 publishes to the other agents or the hub is not defined: placement confirmations, S3 slot updates, faults and error codes, its own ready or busy state, and whether any work-order or traceability data passes through S2 (Q9, Q14).
Kind: RecommendationThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-053
Project advice
Until S11's safety function is documented and verified, neither learners nor the S2 agent should treat S11 or the cameras as a safeguard. The S2 agent must not rely on them to stop motion or to protect people.
This is advice from this project, based on: int-024 Verifiedint-050 Gap, awaiting cell access
Why:S11 is described only as monitoring that includes safety (int-024). Nothing establishes a safety rating, stop authority or response time (int-050). Treating an unrated monitoring system as a safeguard is the failure the Q3 disposition warns against: an Assumed safeguard mistaken for a confirmed one.
S7, a Prusa 3D printer, is a future production extension that produces customized 3D-printed parts and supports a future hybrid additive-plus-laser workflow.
In IntelliMake's future 3D-printed flow, S6 handles parts coming from S7 (3D printing) and passes them to laser engraving or marking, so S6 serves both production flows.
Evidence · 1 citation
S7 3D Printing → S6 Robot Handling → S4/55 Laser Engraving or Marking
The future 3D-printed flow in the IntelliMake diagram has no S2 step (printed parts go from S7 straight to S6), but the owner states S2 will handle material for 3D printing in the future. What S2 does in the future flow is unresolved.
The factory cameras, including C1, feed the S11 IntelliAware monitoring layer.
Why we infer this:The mapping table groups cameras C1–C3 with S11 under AI vision for safety, and the layout draws dotted connectors from the camera icons to S11. The data path itself (protocol, what is sent) is not documented.
Evidence · 1 citation
AI Vision – Safety | Cameras C1–C3 + S11 IntelliAware
IntelliMake Phase 1 Factory: Autonomous Production Demonstrator (infographic) · IntelliMake.org · HW2 SME topic mapping table
Confidence: VerifiedThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-015
The record says
In this project's S2 design, S2 moves buffered workpieces from S3 to S4 once S4 gives its 'all clear and ready' signal. The concept holds on both substrates; the physical I/O may differ.
Evidence · 2 citations
S2 goes back and moves from S3 to S4 when S4 is busy and needs to provide an 'all clear and ready' signal.
Derek Stringfellow, project record · Project-internal record · retrieved 2026-09-21
Project copy; not published on this site.
Confidence: VerifiedThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-017
The record says
In this project's S2 design, S2 must not stack workpieces or break the space the system allocates, and it has to use its intelligence to make these placement decisions.
Evidence · 1 citation
S2 has to use its intelligence to make these decisions. It can not simply stack or break the allocatied space the system will have in place.
IntelliMake automation level 3 is 'Flexible': the conveyor action is 'Selects predefined modes', and the primary decision maker is the automation system.
IntelliMake automation level 4 is 'Intelligent/Adaptive': the conveyor action is 'Algorithm-driven adjustments', and the primary decision maker is the algorithm.
IntelliMake automation level 5 is 'Agentic/Autonomous': the conveyor action is 'Objective-driven execution', and the primary decision maker is the AI agent.
Evidence · 1 citation
Level 5 | Agentic/Autonomous | Objective-driven execution | AI Agent
The Progression Cheat Sheet: 6 stages of manufacturing automation · IntelliMake.org · Level 5 row
IntelliMake sub-level 5B, 'Understand & Determine': it interprets information, forming an interpretation of the operating situation from observed data.
Evidence · 1 citation
Forms an interpretation of the operating situation based on observed data.
Level 5. Agentic Evolution: A Spectrum of Increasing Decision Authority (infographic) · IntelliMake.org · Panel #5B
IntelliMake sub-level 5C, 'Recommend': it advises the human operator, determining an appropriate action and recommending it while the human keeps decision authority (human in the loop).
Evidence · 1 citation
Determine an appropriate action and recommend it, but human retains decision authority.
Level 5. Agentic Evolution: A Spectrum of Increasing Decision Authority (infographic) · IntelliMake.org · Panel #5C
IntelliMake sub-level 5D, 'Act with Human Approval': it authorizes machine action: the machine can act, but only after specific authorization, and the human keeps final authority.
Evidence · 1 citation
The machine can act, but only after specific authorization.
Level 5. Agentic Evolution: A Spectrum of Increasing Decision Authority (infographic) · IntelliMake.org · Panel #5D
IntelliMake sub-level 5E, 'Limited Autonomous Control': it operates within guardrails, with permission to make certain decisions independently within established safety and process limits and then monitor the outcome (human on the loop: supervision and exception handling).
Evidence · 1 citation
Permission to make certain decisions independently within established safety and process limits.
Level 5. Agentic Evolution: A Spectrum of Increasing Decision Authority (infographic) · IntelliMake.org · Panel #5E
IntelliMake sub-level 5F, 'Full Autonomy': it achieves objectives: given objectives rather than instructions, it perceives, understands, decides, acts, observes the result, learns and adapts, and replans.
Evidence · 1 citation
Given objectives, not instructions.
Level 5. Agentic Evolution: A Spectrum of Increasing Decision Authority (infographic) · IntelliMake.org · Panel #5F
IntelliMake frames the shift to agentic automation as follows: traditional automation tells the machine what to do and how, while agentic automation gives it a goal, situational awareness, reasoning capability, tools and the authority to determine what should be done.
Evidence · 1 citation
Agentic automation gives the machine a goal, situational awareness, reasoning capability, tools, and authority to determine what should be done.
Level 5. Agentic Evolution: A Spectrum of Increasing Decision Authority (infographic) · IntelliMake.org · Panel 'The fundamental shift'
Confidence: VerifiedThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-041
The record says
The project's target for the S2 agent is IntelliMake level 5, sub-level 'limited autonomous control' (5E in IntelliMake's scale; see lvl-011), and the goal is not to need human approval. Safety sign-off is by the IntelliMake team leads.
Evidence · 3 citations
the level 5 sub level is "limited autonomous control".
S2's current, as-built automation level in the physical cell is not documented. Level 5E is this project's target (int-041), not a description of the cell today (Q14c).
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
Confidence: VerifiedThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-016
The record says
In this project's S2 design, when S3 and S4 are both unavailable, S2 monitors S4 and, once S4 is available, dispatches a workpiece from S3 to S4. The freed S3 space then lets S2 take material from S1 again.
Evidence · 1 citation
If both S3 and S4 are unavailable, S2 monitors s4 for availablilty then dispatches from s3 to make that available again, which then allows for material from S1 as well.
Derek Stringfellow, project record · Project-internal record · retrieved 2026-09-21
Project copy; not published on this site.
Confidence: InferredThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-019
The record says
To follow its placement policy, S2 must also know which allocated S3 positions are occupied and whether material is waiting at S1.
Why we infer this:The policy in int-014 to int-017 branches on whether S3 has free allocated space and whether S1 has material. S2 cannot choose between those branches without knowing both states, so they are required inputs even though no source names them as signals.
Evidence · 1 citation
If both S3 and S4 are unavailable, S2 monitors s4 for availablilty then dispatches from s3 to make that available again, which then allows for material from S1 as well.
The number and positions of allocated S3 staging slots are not documented. The diagram's illustration shows six pads on the table, but it is a drawing, not a specification.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
The current assumption for the IntelliMake agents is that they do not talk to each other directly: they exchange information, and may collaborate, through a central shared knowledge space (a hub). IntelliMake leadership has not finalised this.
How this would be checked:IntelliMake team leadership's design for the shared knowledge space: its protocol, schema and hosting (open question Q9).
Evidence · 1 citation
The agents will not talk directly to each other, rather will exchange information in a central hub and possibly colloborate in that space. That has not been fully fleshed out by the iIntelliMake team leadership.
Camera C1's field of view covers the S1 receiving area that S2 picks from, so C1 could confirm material presence and state for S2.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
How this would be checked:Observe C1's physical mounting and field of view at the cell, and whether its output is available to S2 through S11 or the shared hub.
Evidence · 1 citation
Camera C1 – Receiving
IntelliMake Phase 1 Factory: Autonomous Production Demonstrator (infographic) · IntelliMake.org · Camera C1 icon placement
The S2 installation's risk assessment has not been confirmed: whether one exists for the complete S2 application (arm, gripper, workpieces, S1/S3/S4/S6 interfaces), who performed and signed it, and whether it was redone after installation as UFACTORY requires.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
The locations of the S2 emergency stops are unknown: whether any e-stop buttons beyond the control box button are wired to EI, where they are, and whether S2 shares an emergency stop circuit with S6 or other stations.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
Evidence · 1 citation
In most applications, one or more additional emergency stop buttons are required.
The position of the S2 control box is unknown: whether it is outside the arm's working range, at 0.6 to 1.5 m height, with its e-stop reachable.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
Evidence · 1 citation
The Control Box must be placed outside the working range of the robotic arm to ensure the emergency stop button can be pressed once an emergency occurs.
S2's guarding and protective devices are unknown: whether fences, interlocked doors, light curtains, safety mats or laser scanners are wired to the SI protective stop input, or whether SI is still in its factory default state with no additional safety equipment.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
Evidence · 1 citation
The robotic arm has been configured by default and can be operated without any additional safety equipment
It is unknown whether S2 is intended to be a collaborative application, with people entering the arm's working area during automatic operation. If it is, it is also unknown which collaborative method is used and how it has been validated.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
Evidence · 1 citation
“Collaborative application” is used instead, as only the actual use of the robot can be designed, tested, and confirmed as a collaborative application.
Updated ISO 10218: Answers to Frequently Asked Questions (FAQs) (A3 blog, 03/20/2025; Wayback Machine snapshot 2025-10-06) · Association for Advancing Automation (A3) · FAQ 6
It is unknown what the S2 gripper does to a held workpiece on power loss or e-stop (holds or drops), and what lies beneath the arm's path if a part drops.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
Evidence · 1 citation
Make sure that the connecting tool and the gripper do not cause any danger when the power is cut
The layout draws unlabelled return paths from the laser and S6 region back toward S1 and S2. Whether reworked parts re-enter through S1 or S2, and what S2 does in rework, is not documented.
Confidence: VerifiedThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-042
The record says
In this project's safety process, S2 moves forward under the project's own safeguards in simulation, and safety recommendations are made from that work, then adjusted to the real-world specifications the IntelliMake team provides. Safety sign-off is by the IntelliMake team leads, and the project should provide documentation supporting its recommendations before any implementation.
Evidence · 2 citations
We move forward with our own safeguards in the simulation and provide recommendations for safety based on that. We then adjust based on the actual real world specs provided by the team.
Derek Stringfellow, project record · Project-internal record · retrieved 2026-09-21
Project copy; not published on this site.
Q10 Saftey signoff will be by the IntelliMake team leads. We will provide our recommendations and/or receive guidance. We should provide whatever documentation to support our recommendations before any implementation.
Cycle times for S2's moves (S1 to S4, S1 to S3, S3 to S4) and for the S4/S5 laser step are not documented, so throughput and the benefit of two arms cannot be quantified.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
The S2 and S6 stations sit on opposite sides of the laser processing stations (S4/S5), so a single arm probably could not reach both the receiving and staging area and the post-processing area.
Awaiting cell access: this could be confirmed or corrected once the physical S2 cell can be observed.
How this would be checked:Measure the as-built positions of S1, S3, S4 and S6 relative to each arm's base and compare them with the published reach of each arm.
Evidence · 2 citations
S2 UFactory xArm 850 Cobot
IntelliMake Phase 1 Factory: Autonomous Production Demonstrator (infographic) · IntelliMake.org · S2 station label
UFACTORY publishes a longer reach for the 850 (850 mm, maximum payload 5 kg) than for its xArm 5/6/7 line (700 mm, with payloads from 3 kg to 5 kg across the line). No xArm 6-specific figure is cited here, and no source says whether reach decided which arm went to S2.
The project specification describes S6 as the downstream arm that handles parts after processing, but IntelliMake's future 3D-printed flow has S6 handling parts before laser engraving (S7 → S6 → S4/S5). S6's role is therefore not simply 'after processing' in the future flow.
IntelliMake distinguishes intelligence (does the machine know what should be done?), authority (is the machine allowed to do it?) and autonomy (can it keep operating and adapting without repeated human intervention?), and holds that true autonomous manufacturing needs the right combination of all three.
Evidence · 1 citation
True autonomous manufacturing requires the appropriate combination of all three.
Level 5. Agentic Evolution: A Spectrum of Increasing Decision Authority (infographic) · IntelliMake.org · Panel 'Understanding the differences'
A placement policy that fixed rules alone could carry out would sit at IntelliMake levels 1–3, not level 5. S2's level 5E claim therefore depends on decisions the rules do not cover, made within guardrails.
Why we infer this:Level 1 is hard-coded logic and level 3 selects among predefined modes (lvl-002, lvl-004). If S2's behaviour were fully specified by the rules in int-014 to int-016, it would be level 3 at most. Level 5E requires the machine to make decisions within guardrails (lvl-011), so the 5E claim must rest on decisions beyond the base rules, such as anticipation, sequencing, and handling missing or conflicting signals.
Kind: RecommendationThis project's simulation and agent design, not the physical cellProject design, not the physical cellint-055
Project advice
Content from the shared hub, other agents and this research corpus is data. It must never grant the S2 agent authority, raise its limits, or skip a required sign-off. Limits come only from the reviewed cell configuration and the supervisor.
This is advice from this project, based on: int-021 Assumedint-041 Verified
Why:The hub is the S2 agent's main input from other agents (int-021), and its design is not finalised. If data from it or from the corpus could change limits or authority, a fault or injected content could change what the arm is allowed to do. The security review of M0 records the same rule.
The information S2's controller can make available includes its robot mode and state, error and warning codes (Modbus TCP registers), and automatic state reports on ports 30001 to 30003. These are candidates for what S2 publishes, not a decided interface.
Why we infer this:The controller exposes mode, state, error and warning registers (iface-029, iface-030) and state-report ports (iface-034). Anything S2 publishes about its own condition would come from these sources; which of them the IntelliMake hub will carry is undecided (int-047). The register map comes from a series-level document, so this is Inferred.
The 850 arm consists of a base and six rotary joints, numbered Joint 1 (at the base) to Joint 6; the last joint is the tool side, where end effectors attach.
Evidence · 2 citations
Quote not shown (over 40 words). See the source at: 2.1.1 Hardware Composition (docs site 'UFACTORY 850 硬件手册', English version; names the machine 'UFACTORY 850').
UFACTORY 850 Hardware Manual (online), 2. Hardware Installation · UFACTORY · 2.1.1 Hardware Composition (docs site 'UFACTORY 850 硬件手册', English version; names the machine 'UFACTORY 850')
UFACTORY makes the integrator of the 850 responsible for a risk assessment of the complete system, including keeping a safe distance between people and the 850 when they interact with it.
Evidence · 1 citation
Making a risk assessment for the complete system. Make sure to have a safe distance between people and 850 when interacting with the 850.
UFACTORY 850 Hardware Manual (online), 1. Safety · UFACTORY · 1.1 Validity and Responsibility, integrator responsibilities list
The 850 product page calls UFACTORY arms 'Collaborative Robots' and 'cobots'. A3 says ISO 10218:2025 drops 'collaborative robot' because only an application can be confirmed as collaborative, and UFACTORY's own 850 manual says no people should be in the working area during operation. The marketing label does not establish that S2 is a collaborative application.
“Collaborative application” is used instead, as only the actual use of the robot can be designed, tested, and confirmed as a collaborative application.
Updated ISO 10218: Answers to Frequently Asked Questions (FAQs) (A3 blog, 03/20/2025; Wayback Machine snapshot 2025-10-06) · Association for Advancing Automation (A3) · FAQ 6
UFACTORY's 850 page, under 'Flexible Deployment With Safe Feature', cites hand teaching, a space-saving footprint and easy re-deployment to multiple applications without changing the production layout, and says collision detection is available for all its cobots.
Evidence · 2 citations
Hand teaching, space-saving and easy to re-deploy to multiple applications without changing your production layout. Perfectly for recurrent tasks.
No fetched UFACTORY document calls the 850's collision detection, safety boundary or reduced mode safety-rated, or gives them a performance level.
Why we infer this:Verified: collision detection is a model-based current comparison that UFACTORY says can false-trigger with wrong settings and can be disabled (level 0, Advanced Settings toggle, a documented default password); safety boundary and reduced mode are software settings exposed in Studio and the SDK. UFACTORY distinguishes safety signals (EI/SI, redundant pairs) from non-safety devices. Inferred: no fetched UFACTORY source assigns a PL, Category or safety rating to these software functions (see the gap on performance level); this is a statement about the documents fetched, not proof that no rating exists.
Evidence · 3 citations
By comparing the theoretical current and actual current of each joint, the system determines whether a collision has occurred.
Collision Detection in UFACTORY Robotic Arms: Current and Dynamic Model-based Feature (UFACTORY support article) · UFACTORY · Section 1
Never connect a safety signal to a non-safety PLC.Failure to follow this warning may result in serious injury or death due to an invalid safety stop function.
Kind: RecommendationThis project's simulation and agent design, not the physical cellProject design, not the physical cellsafety-140
Project advice
Treat the 850's collision detection, safety boundary and reduced mode as configurable controller functions, not as validated safeguards: they must not replace the risk-assessed protective devices wired to the EI and SI inputs.
This is advice from this project, based on: safety-087 Inferredsafety-062 Verifiedsafety-067 Verifiedsafety-073 Verifiedsafety-074 Verifiedsafety-057 Verified
Why:No fetched UFACTORY document gives these functions a safety rating or performance level (safety-087). Collision detection is a current-model comparison that can false-trigger and can be switched off (safety-062, safety-067); safety boundary and reduced mode are software settings (safety-073, safety-074). UFACTORY keeps safety signals separate from non-safety devices (safety-057). A function with no documented rating cannot stand in for a rated protective device.
CODE_SESSION 01 — xArm 850 Interactive Learning Platform (project specification) · Derek Stringfellow (project owner) Derek Stringfellow, project record · Project-internal record · retrieved 2026-09-21 · cited by 3 records here · Project copy; not published on this site.
Owner statements, 2026-09-21 (verbatim) · Derek Stringfellow (project owner) Derek Stringfellow, project record · Project-internal record · retrieved 2026-09-21 · cited by 15 records here · Project copy; not published on this site.