Four exact components. Thirty interface checks. Zero validated stacks.

The direct answer

Can the D455 and MID-360 send data to the J4012? Their documents expose plausible, standards-based routes. D455 is a USB 3.1 Gen 1 peripheral and J4012 publishes USB 3.2 Gen 1 host ports. MID-360 publishes 100BASE-TX Ethernet and J4012 publishes 10/100/1000 Ethernet. Those are the audit’s two documented matches.

Can the RRC2054-2 power the stack? Not as a drop-in conclusion. Its 14.40 V nominal and 16.80 V maximum-charge points sit inside the J4012’s 12–24 V input interval and the MID-360’s 9–27 V interval. But the selected battery datasheet does not publish its minimum loaded output or CUV trip voltage, and the stack lacks a complete power budget, connector design, source management, cold-start trace and low-state-of-charge test.

Will the camera and LiDAR be synchronized? The evidence does not establish that. MID-360 supports PTPv2 or GPS / PPS. D455 has its own USB and frame-sync behavior. No selected source defines a common clock bridge or measures the temporal offset between their acquisition timestamps on this J4012.

Is this a validated robot stack? No. RCI found or produced zero exact-pair tests and zero complete four-product validation packages for this release. A documented interface match is a design input, not a compatibility certificate.

The exact four-product boundary

• RealSense D455 — USB stereo-depth camera with RGB and IMU.
• Livox MID-360 — 100BASE-TX panoramic LiDAR with PTPv2 / GPS synchronization inputs and separate 9–27 VDC power.
• Seeed reComputer Industrial J4012 — Jetson Orin NX 16 GB system exposing USB, two Ethernet ports and a 12–24 VDC terminal input.
• RRC2054-2 — 14.40 V nominal smart battery with five blade contacts and SMBus.

The audit does not substitute a developer kit, a different J401 carrier, a D455f, a MID-360S or another RRC20xx pack. Product-family similarity is not exact-model evidence.

Five paths across six compatibility layers

Across all 30 checks, only 2 are direct documented matches. 8 are conditional, 5 require an adapter or harness, 2 are not direct paths, and 9 remain unresolved. The four not-applicable cells prevent battery-only power paths from being penalized for sensor clock or driver fields they do not need.

D455 → J4012: the strongest path still needs a real pair test

The D455 data route is the least ambiguous. RealSense specifies USB 3.1 Gen 1 and Seeed specifies USB 3.2 Gen 1 host ports. The connector shapes differ, so the implementation needs a certified SuperSpeed Type-A-to-Type-C cable rather than a charge-only lead.

RealSense publishes 3.46147 W total power at its stated maximum operating mode. Seeed publishes a 2.1 A maximum shared output-current limit for its stacked USB pair. Dividing the RealSense figure by 5 V gives approximately 0.69 A, which is a useful design comparison but not a measured margin. Another device on the shared pair, cable drop, enumeration transients and projector behavior can change the result.

The software route is also conditional. The RealSense SDK 2.0 beta v2.58.3 support table lists D455 USB with firmware 5.17.3.10 or later and SDK v2.58.3 or later, and names JetPack 5.0.2, 6.0, 6.1, 6.2 and 7.0. The living D400 firmware table separately associates firmware 5.17.3.10 with SDK 2.58.1. RCI preserves both official statements rather than silently choosing one. The Jetson guide validates named NVIDIA board families and versions; none of these surfaces proves an exact J4012 SKU × carrier image × kernel × stream workload. The pair must still be version-frozen and tested.

MID-360 → J4012: Ethernet matches; power does not come through that match

The MID-360 manual specifies 100BASE-TX Ethernet, UDP and a default static-IP workflow. The J4012 publishes two 10/100/1000 Ethernet ports. That establishes a data-link route after subnet configuration.

It does not establish power compatibility through Ethernet. MID-360 uses a 12-pin M12 connector and the Livox splitter separates RJ45 data from bare DC power. J4012 LAN1 is a PoE PSE port, but the MID-360 is not documented as an IEEE 802.3af powered device. A system design should use a reviewed network topology and a separately fused 9–27 V branch sized for the published 18 W startup peak and cold self-heating behavior.

Livox ROS Driver 2 supports MID360 and ROS / ROS 2. J4012 runs JetPack Linux. That is a plausible software route, but no official exact release matrix binds the J4012 carrier image, ROS distribution, Livox SDK / driver and sensor firmware together.

The battery paths show why voltage overlap is not drop-in compatibility

RRC2054-2 nominal and fully charged voltages fall inside both load input intervals. For MID-360 alone, its 18 W startup peak corresponds to roughly 1.25 A at the pack’s 14.40 V nominal point, below the pack’s published 10.0 A maximum-discharge figure. That calculation does not include a J4012 load, low-voltage behavior, cable drop, repeated inrush or temperature.

The connector boundary is decisive. The pack exposes +, C, D, T and − blade contacts. J4012 uses a two-pin terminal input; MID-360 uses M12 power pins. A reviewed system needs the correct RRC mating connector, fuse, wire gauge, DC distribution and preferably an explicit power-management design. Discarding the C, D and T contacts also discards the documented SMBus and temperature path.

The selected RRC datasheet publishes CUV protection but not the trip voltage. Because J4012’s minimum input is 12 V, two in-range points—14.40 V nominal and 16.80 V maximum charge—cannot prove operation through the usable discharge window. Brownout and safe shutdown must be measured.

A common data host is not a common clock

MID-360 supports IEEE 1588-2008 PTPv2 over UDP/IPv4 or GPS / PPS. The selected J4012 pages do not state its NIC hardware-timestamp capability, selected PTP profile or achieved error. D455 hardware sync and USB timestamps are a different domain. Receiving both streams on one computer does not make their acquisition times equivalent.

A defensible stack report must name the master clock, sensor timestamp source, host clock discipline, offset estimator and measurement instrument. It must publish offset, drift and jitter—not only a screenshot showing two ROS topics arriving.

The minimum package that would turn this hypothesis into evidence

1. Freeze identities: J4012 SKU, D455 and MID-360 serial / hardware identity, RRC pack part number, cable and connector part numbers, JetPack / L4T, kernel, drivers and firmware.
2. Publish the wiring: USB port, Ethernet port, static IP, M12 splitter or production harness, RRC mating connector, fuse, conductor gauge, grounding and power-distribution diagram.
3. Instrument power: cold start, sensor startup, maximum compute workload, steady state, low state of charge and controlled shutdown with per-branch voltage and current traces.
4. Measure data integrity: USB enumeration speed, depth / RGB / IMU frame loss, Ethernet packet loss, point-cloud loss, CPU / GPU load, receive buffers and storage latency under concurrent operation.
5. Measure timing: define the master clock and record D455-to-MID-360 offset, drift and jitter across the intended session length.
6. Soak and restart: publish temperatures, throttling, disconnects, brownouts, service restarts and recovery behavior rather than reporting only one successful boot.

11 official source surfaces

Every result is tied to an exact location rather than a search snippet. The audit uses the current RealSense D400 datasheet, v2.58.3 beta release, D400 firmware table and Jetson guide; the Livox MID-360 manual and ROS driver; two Seeed J4012 integration pages; and three RRC battery / connector / power-management surfaces.

Download the thirty observations

Download the immutable robot-stack interface JSON and observation-level CSV. Stable aliases are current JSON and current CSV. Each row carries its path, layer, state, evidence basis, statement, boundary, required action, source IDs and source URLs.

Suggested citation: Robot Component Index. “Can Four Robot Components Form One Stack? 30 Interface Checks, 0 Full-Stack Validations.” RCI 024, research dataset v0.2.0, 2026-08-10. https://robotcomponentindex.com/research/exact-model-robot-stack-interface-audit/

Use the methodology for evidence-state definitions, the six-layer compatibility guide for the general model, and Cite RCI for version-aware machine citation records.