The direct answer
Where did the catalog grow? Between the immutable v0.4.7 release (2026-08-10) and the current v0.4.14 release (2026-08-13), the catalog added 27 exact component records — 76 to 103 — and 19 manufacturers — 26 to 45. All 346 added claim IDs sit on the 27 new components; zero claims were added to, modified on or removed from the 76 pre-existing records. Motion and actuation received the most additions with 8 records, led by four ROBOTIS DYNAMIXEL servo models plus a Harmonic Drive gear unit, a Kollmorgen frameless motor, a maxon flat motor and a Nabtesco cycloidal reducer.
Did evidence quality improve? No — evidence volume grew, evidence quality did not. Of the 346 added claims, 290 entered as manufacturer-claimed (M) and 56 as explicit unknowns (U). Zero entered as independently verified: V stayed at 6, RCI-derived stayed at 1 and conflicting stayed at 6. A catalog that is 35.5 percent larger contains exactly as many independently verified values as before.
What was preserved? Every one of the 599 claim IDs in v0.4.7 appears in v0.4.14 with identical serialized metadata. Zero claim IDs were removed, zero were modified and the six published comparisons carried over unchanged.
27 new records and 346 new claims measure coverage, not correctness. RCI has run zero physical tests on any of the 103 components, and no third party verified a new claim in this window. The added records widen what can be checked; they do not check it.
Where the 27 records landed across five hardware layers
Every added record carries a category field in the frozen JSON, so the layer split below is a recomputation, not an editorial grouping. Deltas sum exactly: 8 + 7 + 7 + 3 + 2 = 27 components and 109 + 93 + 84 + 33 + 27 = 346 claims.
| Hardware layer | Components | Added | Claims | Unknowns (U) | Added records |
|---|---|---|---|---|---|
| Motion and actuation | 18 → 26 | +8 | 147 → 256 (+109) | 8 → 24 | ROBOTIS DYNAMIXEL XL330-M077-T, XL330-M288-T, XL430-W250-T and XM430-W210-T; Harmonic Drive CSF-14-50-2UH; Kollmorgen TBM2G-07608-A-NNAA-00; maxon EC 20 flat iE (688690); Nabtesco RV-20E, ratio 57 |
| Compute and control | 17 → 24 | +7 | 124 → 217 (+93) | 4 → 18 | AMD Kria K26 SOM Commercial; Beckhoff CX5130-0100; Intel Atom x7425E Processor; PJRC Teensy 4.0; Raspberry Pi 5; Raspberry Pi 5 Model B Rev 1.1; ROBOTIS OpenCR 1.0 |
| Perception and sensing | 21 → 28 | +7 | 163 → 247 (+84) | 1 → 15 | ams OSRAM AS5040; Raspberry Pi Camera Module 3 NoIR; RoboSense M2; SLAMTEC RPLIDAR A1M8; TDK InvenSense MPU-6500; Teledyne FLIR Blackfly S BFS-PGE-161S7C; Zivid 2+ M60 |
| Interfaces and integration | 8 → 11 | +3 | 66 → 99 (+33) | 3 → 11 | Moxa EDS-2008-EL; PEAK-System PCAN-USB FD (USB-A); Vector Informatik VN1630A |
| Power and energy | 12 → 14 | +2 | 99 → 126 (+27) | 5 → 9 | Inventus Power S-2S1P-18650; Vicor DCM3623T50M31C2T00 |
Two additions are deliberate near-duplicates: Raspberry Pi 5 and Raspberry Pi 5 Model B Rev 1.1 are separate records because a board revision can change behavior that a family-level page hides. The layer counts describe where cataloging effort went in a three-day window; they say nothing about market size or engineering importance.
The evidence mix: M and U grew, V, D and C did not move
| State | v0.4.7 | v0.4.14 | Delta | Interpretation |
|---|---|---|---|---|
| M · Manufacturer claimed | 565 | 855 | +290 | All 290 added M records cite an official page, e-manual or datasheet for one exact model. They remain claims, not verified values. |
| V · Independently verified | 6 | 6 | +0 | Unchanged. No added record met the independent-verification bar in this window. |
| D · RCI derived | 1 | 1 | +0 | Unchanged. The single explicit RCI derivation carried over untouched. |
| C · Conflicting | 6 | 6 | +0 | Unchanged. No new cross-source conflicts were recorded in this window. |
| U · Unknown | 21 | 77 | +56 | The fastest-growing state. All 56 added U records preserve explicit gaps in public evidence instead of filling them. |
Because zero existing claims changed state, the deltas decompose cleanly: the entire +290 in M and the entire +56 in U come from the 27 new components. The share of manufacturer-claimed records fell slightly, from 94.3 to 90.5 percent, but only because unknowns grew faster — not because verification grew at all. Independently verified claims now cover 0.6 percent of the catalog, down from 1.0 percent, purely by dilution.
Why unknown grew fastest: gaps are recorded, not filled
U was the fastest-growing evidence state in relative terms — 21 to 77, a factor of 3.7 — and this is a policy outcome, not an accident. When RCI checked the cited official documentation for a new record and a field was not published for that exact model, the field was entered as an explicit unknown rather than filled with a family-level figure, a marketing number or a plausible estimate. Three concrete cases from the frozen v0.4.14 file:
- ROBOTIS DYNAMIXEL XM430-W210-T, continuous torque. The record carries manufacturer stall-torque references (2.7 N·m at 11.1 V up to 3.7 N·m at 14.8 V) with ROBOTIS's own warning that stall torque differs from continuous output — and a separate U claim stating that a continuous-torque envelope is not published in the cited public evidence. The same pattern applies to the XL330-M077-T, XL330-M288-T and XL430-W250-T records.
- Vector Informatik VN1630A. Four of its eight claims are unknowns, including timestamp accuracy and the exact channel configuration — fields an integrator would need but that the cited public evidence does not state for this exact interface.
- Harmonic Drive CSF-14-50-2UH, thermal envelope. The rated torque of 5.4 N·m is recorded as a manufacturer claim with its L10 rating condition, while the robot-joint thermal envelope is recorded as unknown rather than inferred from the rating table.
The 56 added unknowns also encode a structural fact: every one of the 27 new components carries exactly one independent-validation field marked U — 27 of the 56 — an explicit record that no independent test result was found for that exact model. The remaining 29 unknowns are unpublished specification fields like the cases above. A rising U count therefore signals honesty about missing evidence, not deteriorating data.
What did not change, stated as boundaries
- 0 removed claim IDs. All 599 v0.4.7 claims persist in v0.4.14 with byte-identical serialized metadata. Zero removals is a revision-control observation for this window, not a promise; any future removal must appear in a versioned delta with a reason.
- V unchanged at 6. The independent-verification layer did not grow while the catalog grew 35.5 percent. Adding manufacturer-sourced records is faster than earning verification, and this brief does not conflate the two.
- Comparisons unchanged at 6. No new cross-product comparison was published against the 27 added records in this window.
- 0 RCI physical tests. RCI has not operated, measured or benchmarked any of the 103 cataloged components. Every M record remains a manufacturer claim.
v0.4.7 remains an immutable 76-component, 599-claim baseline at its frozen URL. RCI 047 publishes the v0.4.7 → v0.4.14 delta instead of silently rewriting the earlier release — the same rule RCI 023 applied to the v0.4.0 → v0.4.5 window.
Method and limits
RCI compared the complete component and claim arrays in the frozen public v0.4.7 catalog and v0.4.14 catalog by stable ID. A component or claim ID present only in v0.4.14 is added; an ID present in both with different serialized metadata would be modified (observed: zero); an ID present only in v0.4.7 would be removed (observed: zero). Evidence-state and layer tables are recomputed from the claim-level evidence codes and the component-level category field. Anyone with both files can reproduce every number on this page.
- This is a reproducible comparison of two published releases — the same recomputation work level as RCI 045's directory delta — not hardware, firmware or SDK testing.
- Manufacturer-claimed values remain claims even when the citation is an official datasheet; inclusion in the catalog is not endorsement or verification.
- The window covers three days of cataloging. Layer and manufacturer deltas describe editorial coverage decisions inside that window and must not be read as market signals.
- Percentage shares are reported to one decimal from exact counts; all other figures are exact.
Download and reproduce Data Brief 003
Both compared releases are frozen and public: v0.4.7 catalog JSON with its claim-level CSV, and v0.4.14 catalog JSON with its claim-level CSV. The current-release aliases are components.json and components.csv; all machine-readable endpoints are listed on the data page.
Suggested citation: Robot Component Index. “Robot Component Catalog Delta: 27 Added Components, 346 Added Claims, 0 Removed IDs.” RCI 047 / Data Brief 003, 2026-08-18. https://robotcomponentindex.com/research/catalog-release-delta-v0-4-7-to-v0-4-14/
For evidence-state definitions, use the methodology. For the previous catalog delta, use RCI 023 / Data Brief 002. For version-aware BibTeX and machine citation records, use Cite RCI.