The first physical path is prepared, not executed
RealSense D455 to Seeed reComputer Industrial J4012 is the first exact-model path. The protocol covers exact identity and software freeze, repeated SuperSpeed enumeration, concurrent depth/RGB/IMU delivery, USB VBUS voltage/current, timestamp offset/drift/jitter, thermal soak and restart behavior.
two evaluation-loan requests submitted; zero loan approvals, specimens, runs or results. A loan request, protocol draft, photograph or successful demo is not a completed bench result.
The internal execution contract is RCI024-BENCH-1. It preserves exact identities, software hashes, predeclared thresholds, every attempted run, raw logs, calibration state, deviations and file hashes. See the EV0–EV6 methodology.
The Wave 1 design was frozen before specimens or results
RCI registered the first D455 → J4012 study design before possessing either exact specimen, booking an instrument or observing a result. It fixes 30 enumeration attempts, a one-hour concurrent stream and timestamp observation, a two-hour thermal run, explicit frame denominators, zero USB 2 fallback and zero SDK restart screening rules.
The registered file records 0 specimens possessed, 0 runs started and 0 results available. If the received exact device does not support a registered mode, RCI must retain the unsupported evidence and issue a new version—not substitute a more favorable mode in place.
Download the machine-readable Wave 1 pre-registration. Its source file SHA-256 is 74e297450d9d8f03f04e7355fce9e618bc845a8042c10930a93eca248a486bf1.
Four collaboration tracks with different evidence meaning
| Track | Possible level | Required boundary |
|---|---|---|
| Protocol review RCI or a future named operator | no EV upgrade by itself | Advice, equipment recommendations or a successful demonstration do not constitute an external replication. |
| Hosted RCI bench test RCI controls the run and analysis; a host may supply facilities or instruments | EV4 | A host facility is acknowledged, but hosting does not make the result an external independent replication. |
| External independent replication a named external organization runs the protocol on a separately identified specimen | EV5 | RCI may supply the protocol or analysis template but cannot operate the external run or suppress an unfavorable result. |
| Applicable accredited-laboratory test a laboratory whose applicable test or calibration scope is verified for the claimed measurement | EV6 | The laboratory name alone is insufficient. EV6 applies only to the measurements covered by the verified scope and report. |
Protocol advice does not upgrade evidence. A hosted run can be EV4 when RCI operates and documents the test. EV5 requires a named external organization to run the protocol on a separately identified specimen. EV6 additionally requires an applicable accredited scope—not merely a well-known laboratory name.
Editorial independence survives loans, facilities and paid laboratory time
- A donated, loaned or discounted specimen is disclosed with the provider, dates and return conditions.
- A supplier may correct factual identity or safety errors but receives no approval right over protocol, thresholds, findings, headline or publication timing.
- Payment may buy laboratory time or calibration, never a favorable result, evidence level, ranking or omission of failed runs.
- RCI and every external operator disclose employment, funding, equipment, travel, consulting, investment and other material relationships.
- All attempted runs stay in the denominator; exclusions and deviations are dated and justified before summary publication.
- Photographs and video prove identity or procedure only. They do not replace raw measurements, instrument records or uncertainty.
- A coauthored report identifies who designed, operated, analyzed, reviewed and funded each part of the work.
- No organization is described as an RCI partner until both parties have explicitly agreed to the stated scope.
A manufacturer can loan hardware and correct an identity or safety error. It cannot approve thresholds, remove failed runs, choose the conclusion or delay an unfavorable result. Every material relationship is disclosed beside the result.
Candidate organizations, not announced partners
These organizations have public capabilities relevant to one part of the program. RCI sent bounded enquiries to the first three candidates on 2026-08-11 and is awaiting replies. Delivery does not establish interest, capability or partnership; RCI has not obtained availability, pricing, specimen access, a named operator or an agreement from any candidate.
| Candidate | Possible role | Current boundary |
|---|---|---|
| A*STAR ARTC / ROS-Industrial Consortium Asia Pacific Singapore · contacted-awaiting-reply | protocol review, robotics testbed or hosted EV4 engineering collaboration | Its industry and robotics collaboration role is relevant, but no facility, staff, schedule, price or independent-replication commitment has been agreed. |
| NTU Robotics Research Centre Singapore · contacted-awaiting-reply | academic protocol review or separately operated EV5 replication | Robotics and perception capability makes it a plausible candidate; no researcher, laboratory, specimen or replication has been committed. |
| A*STAR National Metrology Centre Singapore · contacted-awaiting-reply | instrument calibration, measurement uncertainty or traceability support | Calibration can strengthen EV4/EV5 measurement traceability but is not by itself a robot-component performance replication or EV6 result. |
| TÜV SÜD Singapore Singapore / international · candidate-not-contacted | safety, conformity or applicable accredited testing after scope confirmation | RCI must verify the exact laboratory, method and accreditation scope before assigning EV6; general robotics certification capability is not a D455 accuracy or J4012 performance result. |
| UMass Lowell NERVE Center United States · candidate-not-contacted | later-stage robotics capability benchmarking or EV5 replication | Its robotics test environment is relevant to later system-level validation; distance, cost and exact component-level fit remain unassessed. |
What a collaborator must be willing to preserve
- Exact specimen and serial-to-role identity, including firmware and software.
- Named operator, facility, instrument, calibration and uncertainty records.
- A frozen protocol and thresholds before the first measurement.
- Complete raw files, failed attempts, exclusions, deviations and analysis code.
- A clear funding, equipment-loan and conflict disclosure.
- Permission to publish the bounded result, including an unfavorable or non-replicating outcome.
Download the machine-readable program ↓
RCI currently has no announced laboratory, university or testing-center partner and has completed zero runs under the first protocol.