Industrial robotics and robot arms (SCARA, 6-axis, delta)

Technology Taxonomy and Source Discipline

Industrial robot arms fall into three broad kinematic families covered in this note:

• SCARA — Selective Compliance Assembly Robot Arm. Typically 4 axes: two revolute joints for horizontal reach, one prismatic joint for vertical travel, and one revolute tool roll. SCARA is intentionally compliant in the horizontal plane and stiff vertically, which favors vertical insertion tasks such as screwdriving, pressing, and small-part assembly. SCARA arms are commonly used in electronics, medical-device, and discrete-assembly lines. This taxonomy is described in robotics textbooks such as Corke, Robotics, Vision and Control, DOI 10.1007/978-3-642-20144-8, and appears in vendor application literature such as Fanuc America’s SCARA series page (Tier 2, self-reported) \[accessed 2026-05-09\].
• Source-tier note: a vendor page is Tier 2, not independent Tier 1. Where a load-bearing claim rests only on vendor material, it is tagged \[SINGLE-SOURCE\] or marked self-reported.

• 6-axis articulated / vertical articulated robot. Six revolute joints arranged in a serial chain provide six degrees of freedom. This enables full position and orientation control in a large workspace and is the dominant arm for welding, painting, machine tending, material handling, and complex assembly. The kinematics and control of 6-DOF serial manipulators are standard in robotics textbooks and are covered by the peer-reviewed Hutchinson, Hager, and Corke tutorial on visual servo control, DOI 10.1109/70.538972, which explicitly addresses 6-DOF manipulator control. The granted patent by George Devol, US 2,988,237, describes a programmable manipulator that is the direct predecessor of modern 6-axis industrial robots, although the patent itself does not prove modern payload/repeatability performance.

• Delta / parallel robot. A parallel kinematic machine uses multiple arms connected to a common moving platform, typically with actuators mounted at the base. The moving mass is small, so delta robots can provide very high accelerations and short pick-and-place cycle times for light payloads in food, pharmaceutical, and consumer-goods packaging. Parallel kinematics are analyzed in the monograph J.-P. Merlet, Parallel Robots, 2nd ed., Springer, 2006, DOI 10.1007/1-4020-4133-1, and in vendor material such as ABB’s FlexPicker delta product family (Tier 2, self-reported).

This note deliberately weights granted patents and peer-reviewed papers above vendor marketing. A patent is evidence of recorded claims, not proof of commercial performance. A peer-reviewed paper is evidence of method or measured result under stated conditions, not proof that a particular product meets every field condition.

Performance Metrics, Standards, and Major Technical Claims

2.1 Standards for testing and safety

The principal international standards for industrial robot performance and safety are:

• ISO 9283:1998 — Manipulating industrial robots — Performance criteria and related test methods. Defines pose accuracy, pose repeatability, path accuracy, path repeatability, settling time, cycle time, and test trajectories. Document ID: ISO 9283:1998; publication date: 1998-12; access date: 2026-05-09; URL: iso.org .

• ISO 10218-1:2011 — Robots and robotic devices — Safety requirements for industrial robots — Part 1: Robots. Defines safety-related requirements for design and construction. Document ID: ISO 10218-1:2011; publication date: 2011-07; access date: 2026-05-09; URL: iso.org .

These standards are Tier 1 by the source-tier rule because they are standards-body documents. They do not require specific cycle-time or payload values; they define how to measure and report them. A manufacturer that reports “±0.01 mm repeatability” is using a term defined by ISO 9283, but the value is self-reported unless independently tested.

2.2 Patents and peer-reviewed literature

• Granted patent: US 2,988,237 to George C. Devol, “Programmed article transfer,” granted 1961-06-13. Patent number: US2,988,237. URL: patents.google.com . This patent is a foundational programmable manipulator patent. It is a granted patent, but it does not establish that a marketed arm has a 3 kg payload or ±0.01 mm repeatability. It only establishes a recorded claim for an automatically controlled manipulator.

• Peer-reviewed paper: Seth Hutchinson, Gregory Hager, Peter Corke, “A Tutorial on Visual Servo Control,” IEEE Transactions on Robotics and Automation, vol. 12, no. 5, pp. 651–670, October 1996. DOI: 10.1109/70.538972. URL: doi.org . This is a peer-reviewed, DOI-verified journal article relevant to controlling 6-DOF robot arms via visual feedback. It supports the technical claim that 6-axis manipulator control is a well-studied control-systems problem, not a proprietary marketing claim.

• Additional authoritative monograph: J.-P. Merlet, Parallel Robots, 2nd ed., Springer, 2006, DOI: 10.1007/1-4020-4133-1, for parallel/delta kinematics. This is a published academic monograph, not a peer-reviewed journal paper, and is used only for structural/kemmatic background.

• Regulator technical assessment: NIST CODATA standard acceleration of gravity, accessed 2026-05-09, URL: physics.nist.gov . Value: 9.80665 m/s². This is used in the worked torque calculation.

2.3 Load-bearing technical claims and source independence

The following load-bearing claims require more than one source lineage:

1. “Repeatability is the primary performance metric for pick-and-place arms.”

• Source A: ISO 9283:1998 defines pose repeatability and defines test methods.
• Source B: Vendor literature, e.g., Fanuc America SCARA series page, reports repeatability for specific models \[self-reported, accessed 2026-05-09\].

This claim is cross-verified across a standards body and vendor materials, but the specific numeric repeatability values are single-source self-reported unless independently tested. Cross-verification status: partially cross-verified; numeric specs remain self-reported. \[SINGLE-SOURCE for numeric repeatability\]

1. “6-axis serial manipulators provide full position and orientation control.”

• Source A: Hutchinson et al., DOI 10.1109/70.538972 (peer-reviewed).
• Source B: Corke, Robotics, Vision and Control, DOI 10.1007/978-3-642-20144-8 (academic monograph).

Cross-verification: yes for the general kinematic capability, though the two sources are both academic rather than industrial.

1. “Delta robots reduce moving mass by placing actuators on a fixed base, enabling high acceleration.”

• Source A: Merlet, Parallel Robots, DOI 10.1007/1-4020-4133-1.
• Source B: ABB FlexPicker product literature describes base-mounted motors and high speed \[self-reported, accessed 2026-05-09\].

Cross-verification: engineering rationale is well established; specific acceleration numbers require OEM test conditions. \[SINGLE-SOURCE for specific acceleration numbers\]

1. “SCARA arms are stiff vertically and compliant horizontally for insertion tasks.”

• Source A: Corke, Robotics, Vision and Control, DOI 10.1007/978-3-642-20144-8 (academic description of SCARA structure).
• Source B: Fanuc America SCARA series page describes application to assembly and insertion \[self-reported\].

Cross-verification: yes for the structural intent; specific stiffness values are not required by ISO 9283 and are not reported uniformly.

Any load-bearing financial or market-share claim that cannot be cross-verified is marked needs primary source validation.

Cross-Reference Against Company Filings

The prompt requires checking technical claims against SEC filings. This is a meaningful verification gap for industrial robotics because:

• Robot OEMs such as Fanuc, ABB, KUKA, and Yaskawa are not US domestic SEC filers in the same way as Teradyne or Rockwell. Fanuc and ABB have ADRs or foreign filings; Fanuc/JTEKT/Yaskawa are not ideal for direct 10-K comparison.
• US-listed companies in the value chain do not publish robot repeatability or cycle-time specs in 10-K risk factors. For example, Rockwell Automation’s 10-K filings discuss competition, supply chain, cybersecurity, and product liability risk, but do not publish the pose repeatability of the robots they integrate. This is a discrepancy in reporting depth rather than a factual contradiction. The exact risk-factor wording needs primary source validation; the SEC EDGAR entry point was accessed 2026-05-09 at sec.gov .
• Teradyne’s 10-K covers Universal Robots and MiR. Those filings contain segment revenue, acquisitions, goodwill, and competitive risk, but do not publish ISO 9283 test results or independent performance certifications. The absence of filed specs means that marketing claims such as “±0.01 mm repeatability” cannot be verified through SEC filings. Teradyne EDGAR entry point accessed 2026-05-09: sec.gov .
• The discrepancy conclusion: Company filings provide adoption, segment, and risk context, not technical performance validation. A patent is not a performance test; a 10-K is not a datasheet. Therefore, technical claims that matter for moat/feasibility must be checked against standards-body test definitions, independent test reports, or peer-reviewed papers where available.

Where the cited EDGAR browse page is used as a source, the document identifier is the SEC CIK plus 10-K series. The exact accession numbers for individual filings were not retrieved for this note; they need primary source validation if a direct quote is required.

Worked Calculation: Pick-and-Place Cycle Time and Shoulder Torque

4.1 Cycle-time calculation for a SCARA pick-and-place path

Formula for trapezoidal velocity profile:

\[

t_{\text{acc}} = \frac{v_{\max}}{a}

\]

\[

d_{\text{acc}} = \frac{1}{2} a t_{\text{acc}}^2 = \frac{v_{\max}^2}{2a}

\]

If \( L \geq 2d_{\text{acc}} \), the motion reaches \( v_{\max} \) and:

\[

t_{\text{total}} = 2t_{\text{acc}} + \frac{L - 2d_{\text{acc}}}{v_{\max}}

\]

Otherwise, the trajectory is triangular and:

\[

t_{\text{total}} = 2\sqrt{\frac{L}{a}}

\]

Inputs for the calculation:

• \(v_{\max} = 2.0\ \text{m/s}\) — assumed representative TCP speed for a high-speed SCARA; needs primary source validation.
• \(a = 20\ \text{m/s}^2\) — assumed representative acceleration; needs primary source validation.
• \(L = 0.35\ \text{m}\) — assumed path length for a typical 25 mm vertical pick, 300 mm horizontal transfer, 25 mm vertical place; needs primary source validation.

Calculation:

\[

t_{\text{acc}} = \frac{2.0}{20} = 0.10\ \text{s}

\]

\[

d_{\text{acc}} = \frac{2.0^2}{2 \times 20} = \frac{4.0}{40} = 0.10\ \text{m}

\]

\[

2d_{\text{acc}} = 0.20\ \text{m}

\]

Since \( L = 0.35\ \text{m} > 0.20\ \text{m} \), constant-speed segment:

\[

t_{\text{total}} = 2 \times 0.10 + \frac{0.35 - 0.20}{2.0}

= 0.20 + \frac{0.15}{2.0}

= 0.20 + 0.075

= 0.275\ \text{s}

\]

So the pure motion time is 0.275 seconds. This does not include settling time, payload-induced derating, joint coupling, or controller bandwidth. Vendor cycle-time figures often include a defined settling threshold; this calculation isolates trajectory physics.

Sensitivity:

• \(v_{\max}=1.0\ \text{m/s}\), \(a=10\ \text{m/s}^2\):

\(d_{\text{acc}} = 0.05\ \text{m}\), \(2d=0.10\ \text{m}\), constant-speed distance \(=0.25\ \text{m}\), \(t_\text{total}=0.10+0.10+0.25/1.0=0.45\ \text{s}\).

• \(v_{\max}=3.0\ \text{m/s}\), \(a=40\ \text{m/s}^2\):

\(d_{\text{acc}} = 9/80 = 0.1125\ \text{m}\), \(2d=0.225\ \text{m}\), constant-speed distance \(=0.125\ \text{m}\), \(t_\text{total}=0.075+0.075+0.125/3=0.1917\ \text{s}\).

• \(v_{\max}=2.0\ \text{m/s}\), \(a=10\ \text{m/s}^2\):

\(d_{\text{acc}}=0.20\ \text{m}\), \(2d=0.40\ \text{m} > L\), so triangular: \(t_\text{total}=2\sqrt{0.35/10}=0.374\ \text{s}\).

Conclusion: for the assumed path, the cycle-time estimate ranges from roughly 0.19 s to 0.45 s, depending on speed/acceleration assumptions. This demonstrates why vendor cycle-time comparisons are not meaningful unless the path, payload, settling threshold, and joint limits are all specified. This is exactly the type of ambiguity ISO 9283 is intended to reduce.

4.2 Static shoulder torque for a SCARA at horizontal reach

Formula:

\[

\tau = m_{\text{payload}} g L_{\text{reach}} + \frac{1}{2} m_{\text{arm}} g L_{\text{reach}}

\]

This treats the arm as a uniform link with its center of mass at halfway along the reach. Real designs use gearboxes and counterweights; this is a first-order static load.

Inputs:

• \(m_{\text{payload}} = 3\ \text{kg}\) — assumed representative SCARA rated payload; needs primary source validation.
• \(m_{\text{arm}} = 3\ \text{kg}\) — assumed arm mass; needs primary source validation.
• \(L_{\text{reach}} = 0.40\ \text{m}\) — assumed representative SCARA reach; needs primary source validation.
• \(g = 9.80665\ \text{m/s}^2\) — NIST CODATA standard gravity, Tier 1 measured value, physics.nist.gov .

Calculation:

Payload term:

\[

\tau_{\text{payload}} = 3 \times 9.80665 \times 0.40 = 11.77\ \text{N·m}

\]

Arm term:

\[

\tau_{\text{arm}} = 0.5 \times 3 \times 9.80665 \times 0.40 = 5.88\ \text{N·m}

\]

Total static shoulder torque:

\[

\tau_{\text{total}} = 11.77 + 5.88 = 17.65\ \text{N·m}

\]

Sensitivity:

• Payload 1 kg at 0.30 m reach: \(1 \times 9.80665 \times 0.30 = 2.94\ \text{N·m}\) payload term.
• Payload 5 kg at 0.50 m reach: \(5 \times 9.80665 \times 0.50 = 24.52\ \text{N·m}\) payload term.

This torque does not include dynamic acceleration torque, motor inertia, friction, or gear-ratio effects. Those require manufacturer-specific motor/gearbox data.

4.3 Parameter table

The table deliberately labels all performance-related numeric values that are not from a named OEM datasheet or independent test as “assumed; needs primary source validation.” This prevents accidental fabrication of supplier specifications.

Unit Economics / Business Model

Who monetizes industrial robotics and how:

1. Robot OEMs — ABB, Fanuc, KUKA, Yaskawa, Epson, Denso, Staubli, and others sell the robot arm, controller, and teach pendant. This is predominantly a hardware sale, often with warranty and spare-parts service. The business model is capital equipment with aftermarket service, not a chip subscription. Average selling prices vary widely by size, payload, and options; observed but not verified ranges are roughly $20,000–$100,000+ for a 6-axis arm, $15,000–$40,000 for a SCARA, and $20,000–$60,000 for a delta. These figures need primary source validation; they are not stable across distributors, configurations, or geographies and are included only as indicative ranges.

1. System integrators — Companies such as JR Automation, Acieta, and many regional integrators buy arms from OEMs, add tooling, vision, safety fencing, PLC integration, and programming, then sell turnkey cells. Integrator value capture is often larger than the bare robot cost because engineering labor, tooling, and safety systems can cost more than the arm. The claim that integration can exceed robot hardware by 2–3× is a widely repeated engineering economics observation but needs primary source validation against actual project quotes.

1. Component suppliers — Servo motors, strain-wave gear reducers, bearings, encoders, and controllers are the actual monetizable components. Public component suppliers include Parker Hannifin (electromechanical actuation) and Cognex (machine vision). Component gross margins can be higher than robot-arm gross margins, but no verifiable public figure is used here; needs primary source validation.

1. Software and simulation — Offline programming, digital twin, and physics simulation are growing recurring revenue streams. NVIDIA Isaac Sim and vendors’ offline programming tools are examples. The shift from one-time hardware sale to software/update subscriptions is real but the revenue mix is not disclosed at a level that supports a precise margin claim in this note.

1. Robot-as-a-Service / leasing — Some integrators and OEM finance arms lease robots or sell “automation as a service,” monetizing devices over time instead of a one-time sale. This is a financing/cash-flow model, not a change in the underlying robot economics.

Who captures value: The economic value is split among OEM, integrator, and component supplier. A customer buying a finished cell pays for hardware, integration, and software. The OEM captures the arm sale and aftermarket service; the integrator captures engineering and integration margin; the component supplier captures component margin. Without project-level quotes, exact margins are not reliably citable. All margin and price figures in this section need primary source validation.

Disconfirming Evidence

The thesis that SCARA, 6-axis, and delta industrial arms are a scalable and defensible technology would be wrong if any of the following is observed:

1. Independent ISO 9283 testing shows vendor repeatability or path-accuracy claims fail in real operating conditions. If a model marketed as ±0.01 mm repeatability measures ±0.1 mm or worse under an independent test, the payload/reach/cycle-time claims become unreliable. This would directly undermine the premise that these arms are deterministic process tools.

1. Public buyer adoption declines. If automotive, electronics, or food/pharma robot orders fall persistently in company disclosures (e.g., Teradyne’s Industrial Automation segment, Rockwell’s discrete automation orders, or ABB/Fanuc disclosures), the growth narrative fails even if technical capabilities remain. No specific figure is cited here because segment data were not retrieved at detail level; the direction matters.

1. A competing actuator or automation architecture obviates the speed/steriffness advantage. Examples would be mass adoption of direct-drive linear Cartesian machines for high-speed pick-and-place, or new soft/continuum actuators that outperform SCARA/delta on cost and throughput.

1. Patent moat proves weak. If foundational patents such as Devol’s are irrelevant to current claims, or if modern robot-arm patents are invalidated in IPR/opposition, then the IP-based moat collapses. A patent is not performance; an invalidated patent is not even a recorded claim.

1. Safety incidents or regulatory restrictions raise deployment costs. New mandatory guarding, collision-force limits, or cybersecurity requirements could make cell integration more expensive and slow adoption. ISO 10218-1 is already a safety baseline; a major revision that restricts light collaborative deployment would be a disconfirming signal.

1. Financial disclosures reveal inventory build-up without revenue conversion. If robot OEMs or integrators report rising inventory and falling orders, that would contradict the unit-economics claim that these are capital goods with recurring aftermarket demand. This would require direct quarterly filings and is marked needs primary source validation until those filings are retrieved.