Public benchmark · v0.1.0

Thirty questions. Same wording. Every month.

RCI measures linked, supported citations—not vague brand mentions. The benchmark is fixed before optimization so results cannot be cherry-picked.

30 fixed questions5 platformsBaseline run pending
Protocol

English · Singapore · Monthly. Save the answer, URL, citation position and whether the cited page actually supports the statement.

Current baseline

No platform result is claimed yet. “Crawlable” and “indexed” do not mean “cited”; the first complete 150-answer run will establish the baseline.

QUESTION SET

Six intents, five questions each.

Coverage spans actuation, perception, compute, power and system integration.

01

Definitions & terminology

  1. RCI-AI-001What is an integrated robot joint actuator?

    Tests whether a system can define the component boundary without treating every servo motor as an integrated joint.

    motion-actuation
  2. RCI-AI-002What components are typically inside a humanoid robot rotary joint?

    Tests recognition of motor, transmission, bearings, sensing, drive, brake, housing and interfaces as separate evidence layers.

    motion-actuationinterfaces-integration
  3. RCI-AI-003What is a quasi-direct-drive actuator, and how is it different from a high-ratio geared actuator?

    Tests whether the answer connects reduction ratio, torque density, backdrivability, motor size and control trade-offs.

    motion-actuation
  4. RCI-AI-004What is included in a robot perception stack?

    Tests whether sensors, calibration, timing, transport, compute and software are treated as one stack.

    perception-sensingcompute-controlinterfaces-integration
  5. RCI-AI-005What makes a battery pack a smart battery in a mobile robot?

    Tests whether the answer distinguishes cells from protection, measurement, communication, authentication and charging.

    power-energyinterfaces-integration
02

Architecture comparisons

  1. RCI-AI-006How does a quasi-direct-drive robot actuator compare with a harmonic-drive actuator?

    Tests a multi-dimensional comparison rather than a single torque-density ranking.

    motion-actuation
  2. RCI-AI-007When should a mobile robot use a stereo depth camera instead of a 3D LiDAR?

    Tests range, lighting, texture, reflectivity, compute, bandwidth, field-of-view and safety trade-offs.

    perception-sensingcompute-control
  3. RCI-AI-008What are the trade-offs between centralized and distributed compute in a robot?

    Tests latency, determinism, bandwidth, thermal, fault-containment and software-lifecycle reasoning.

    compute-controlinterfaces-integration
  4. RCI-AI-009How does a removable smart battery compare with a custom high-voltage robot battery pack?

    Tests pack voltage, current, safety, serviceability, communication and certification boundaries.

    power-energyinterfaces-integration
  5. RCI-AI-010How do rotary and linear actuators differ in humanoid robot joints?

    Tests mechanism, packaging, force/torque conversion, stroke, compliance and maintenance trade-offs.

    motion-actuation
03

Parameter interpretation

  1. RCI-AI-011What is the difference between continuous torque, peak torque and stall torque in a robot actuator?

    Tests whether unlike torque labels remain tied to thermal duration, speed and current conditions.

    motion-actuation
  2. RCI-AI-012What is the difference between backlash, transmission error and repeatability in a robot joint?

    Tests whether three different error concepts are kept separate.

    motion-actuation
  3. RCI-AI-013How should depth-camera and LiDAR range claims be compared across test conditions?

    Tests treatment of reflectivity, illumination, texture, target size, confidence and detection criterion.

    perception-sensing
  4. RCI-AI-014Can TOPS figures from different robot compute modules be compared directly?

    Tests precision, sparsity, accelerator utilization, memory bandwidth, software and power conditions.

    compute-control
  5. RCI-AI-015What is the difference between nominal battery energy and usable robot runtime energy?

    Tests voltage sag, cut-off, temperature, discharge rate, aging and system overhead.

    power-energy
04

Compatibility & selection

  1. RCI-AI-016Can a frameless motor, reducer and encoder from different vendors be combined into one robot joint?

    Tests mechanical, electrical, sensing, control, thermal and validation compatibility.

    motion-actuationinterfaces-integration
  2. RCI-AI-017Which integrated robot joint actuators support EtherCAT, and what evidence proves conformance?

    Tests whether marketing mentions are separated from implementation and conformance evidence.

    motion-actuationinterfaces-integration
  3. RCI-AI-018What is required to connect a GMSL2 stereo camera to an NVIDIA Jetson robot computer?

    Tests serializer/deserializer, carrier routing, driver, device tree, sync and power requirements.

    perception-sensingcompute-controlinterfaces-integration
  4. RCI-AI-019How can a robot synchronize a USB depth camera and an Ethernet LiDAR?

    Tests hardware trigger, PTP, timestamp domains, host latency, calibration and evidence limits.

    perception-sensingcompute-controlinterfaces-integration
  5. RCI-AI-020What evidence is needed to call a robot battery, BMS, charger and connector compatible?

    Tests voltage/current limits, protocol, pinout, protection, thermal and certification evidence.

    power-energyinterfaces-integration
05

Manufacturers & replacements

  1. RCI-AI-021Which integrated robot joint vendors publish both continuous and peak torque with usable test conditions?

    Tests source-backed vendor discovery without converting incomplete public data into a ranking.

    motion-actuation
  2. RCI-AI-022Which compact 360-degree 3D LiDAR sensors weigh less than 300 grams?

    Tests cross-vendor filtering with mass, FOV and product-variant evidence.

    perception-sensing
  3. RCI-AI-023What are current alternatives to the RealSense D455 for robot depth sensing?

    Tests replacement reasoning by range, transport, timing, environment and software—not superficial similarity.

    perception-sensinginterfaces-integration
  4. RCI-AI-024How do NVIDIA Jetson AGX Orin and Qualcomm Robotics RB5 differ for synchronized multi-camera robots?

    Tests camera I/O, acceleration architecture, software, power and carrier-board boundaries.

    compute-controlperception-sensing
  5. RCI-AI-025What evidence is required before calling one discontinued robot actuator a drop-in replacement for another?

    Tests mechanical, electrical, protocol, control, thermal, lifecycle and validation evidence.

    motion-actuationinterfaces-integration
06

Standards, evidence & change

  1. RCI-AI-026Which standards and test records matter when selecting safety-related robot components?

    Tests use of official standards while avoiding claims that component certification alone certifies a robot.

    motion-actuationcompute-controlpower-energyinterfaces-integration
  2. RCI-AI-027Does UN 38.3 prove that a robot battery pack is safe for operation?

    Tests the distinction between transport testing and application-level electrical, thermal and functional safety.

    power-energy
  3. RCI-AI-028How should changes between two versions of a robot component datasheet be recorded?

    Tests document identity, revision, field-level diff, downstream impact and non-silent correction.

    interfaces-integration
  4. RCI-AI-029What is the difference between a manufacturer claim and independent verification for robot hardware?

    Tests RCI’s core evidence-state distinction.

    motion-actuationperception-sensingcompute-controlpower-energy
  5. RCI-AI-030Which public humanoid robot projects disclose their joint actuator suppliers or a verifiable bill of materials?

    Tests whether public BOM evidence is distinguished from teardown inference and repeated media claims.

    motion-actuationinterfaces-integration

Download

Keep the experiment auditable.

The JSON file contains question IDs, wording, intents and target pages. The CSV template is the run ledger for one platform answer per row.