{ "title": "CDPR Actuator-Level Policy Evidence: 4-Cable Simulation to One 8-Motor Robot, Zero RCI Runs", "rciNumber": "RCI 043", "version": "0.1.0", "schemaVersion": 1, "released": "2026-08-11", "canonical": "https://robotcomponentindex.com/research/cdpr-actuator-level-policy-evidence/", "scope": "A facts-first review of an actuator-level deep-reinforcement-learning policy for cable-driven parallel robots. It separates training on one simulated four-cable translational point-mass CDPR, transfer to selected eight-cable configurations and one author-reported real eight-motor trajectory from arbitrary-configuration validation, repeatable hardware evidence, component identity and zero RCI execution.", "sourceIdentity": { "title": "Generalizing deep reinforcement learning across cable-driven parallel robot configurations with actuator-level policies", "authors": [ "Abir Bouaouda", "Mohamed Boutayeb", "François Charpillet", "Dominique Martinez", "Rémi Pannequin" ], "halId": "hal-05613978v3", "halUrl": "https://inria.hal.science/hal-05613978v3", "halFileUrl": "https://inria.hal.science/hal-05613978/file/main.pdf", "halSubmittedAt": "2026-07-30T14:29:13Z", "halPdfPages": 10, "halPdfBytes": 4072632, "halPdfSha256": "efe7e67ed0dbc7aa3304d42c529af31a83eaafd048916c2e6d9845383f155394", "halFileLicense": "CC-BY-NC-ND-4.0", "halFileLicenseUrl": "https://creativecommons.org/licenses/by-nc-nd/4.0/", "arxivId": "2608.07546v1", "arxivUrl": "https://arxiv.org/abs/2608.07546v1", "arxivPdfPages": 9, "arxivPdfBytes": 3934595, "arxivPdfSha256": "5b03a55e841427f825d19b03438ef0bb71b4f213158564ce1a636ae6914805dc", "ssrnAbstractId": "6730496", "ssrnDoi": "10.2139/ssrn.6730496", "rightsTreatment": "RCI publishes independently worded factual abstraction and original analysis with attribution to the HAL v3 record. RCI does not mirror the PDF, reproduce or adapt figures/tables/extended text, use licensed expression commercially, or apply HAL's file license to SSRN's all-rights-reserved copy." }, "headlineEvidence": { "trainingCableCount": 4, "reportedTransferCableCount": 8, "configuredControlFrequencyHz": 100, "authorReportedRealPositionErrorNormUpperBoundM": 0.04, "rciSimulationRuns": 0, "rciHardwareRuns": 0, "interpretation": "The fixed paper reports selected simulation and one physical-CDPR result. It does not demonstrate every actuator count, geometry, rotational platform, load, cable behavior or hardware stack." }, "observations": [ { "id": "RCI043-O01", "field": "training_robot_configuration", "value": "simulated four-cable translational CDPR with a point-mass mobile platform and all cables attached to one point", "evidenceClass": "author-reported-method", "boundary": "One training geometry is not a representative sample of all CDPR configurations." }, { "id": "RCI043-O02", "field": "policy_granularity", "value": "one shared actuator-level policy maps an individual actuator state to its motor command", "evidenceClass": "author-reported-method", "boundary": "A shared architecture does not by itself establish configuration-independent performance." }, { "id": "RCI043-O03", "field": "actuator_state_fields", "value": "current cable length, target cable length, cable length at t-3 and motor current", "evidenceClass": "author-reported-method", "boundary": "Signals and actions are normalized; exact sensors, interfaces and runtime implementation are not disclosed." }, { "id": "RCI043-O04", "field": "configured_control_frequency", "value": 100, "unit": "Hz", "evidenceClass": "author-reported-configuration", "boundary": "A configured frequency is not measured closed-loop latency, jitter or deadline compliance." }, { "id": "RCI043-O05", "field": "simulated_nominal_motor_parameters", "value": "6.8 A; 3420 rpm; 0.8 Nm; 0.123 Nm/A; 0.0178 m drum radius", "evidenceClass": "author-reported-configuration", "boundary": "No orderable motor, drive, encoder, winch or cable is identified." }, { "id": "RCI043-O06", "field": "training_repetition", "value": "four simulation runs with different random seeds", "evidenceClass": "author-reported-experiment-design", "boundary": "Four simulation seeds are not four independent physical replications." }, { "id": "RCI043-O07", "field": "simulation_test_duration", "value": "ten episodes of ten seconds", "evidenceClass": "author-reported-experiment-design", "boundary": "No raw episode trajectories or per-episode result files are published." }, { "id": "RCI043-O08", "field": "simulation_error_ranges", "value": "0–0.5 m conventional RL; 0–0.25 m actuator-level policy", "evidenceClass": "author-reported-result", "boundary": "These are paper-reported ranges tied to one comparison, not RCI recomputation or a universal envelope." }, { "id": "RCI043-O09", "field": "simulated_transfer_configuration", "value": "eight-cable translational point-mass CDPR", "evidenceClass": "author-reported-simulation-result", "boundary": "It does not cover arbitrary cable arrangements, rigid-body rotation or cable physics." }, { "id": "RCI043-O10", "field": "real_hardware_configuration", "value": "one eight-motor translational CDPR", "evidenceClass": "author-reported-hardware-result", "boundary": "No complete orderable BOM, exact motor/drive/sensor models, calibration record or robot identifier is disclosed." }, { "id": "RCI043-O11", "field": "real_hardware_tracking_error", "value": "less than 0.04 m in the Figure 9 trajectory trace", "unit": "position-error norm", "evidenceClass": "author-reported-hardware-result", "boundary": "The paper does not define this as RMSE, MAE, a validated maximum or a confidence interval and publishes no raw trace." }, { "id": "RCI043-O12", "field": "training_cost", "value": "DDPG/SAC: 6–10 s per episode, 10,000–30,000 episodes, 16–85 h; PPO: 4.5–8 s, 40,000–80,000 episodes, 50–100 h", "evidenceClass": "author-reported-compute-result", "boundary": "Compute hardware and software identity are absent, preventing platform comparison." }, { "id": "RCI043-O13", "field": "rotational_cdpr_scope", "value": "not validated; described as a future extension", "evidenceClass": "author-reported-limitation", "boundary": "Reported experiments are restricted to translational point-mass CDPRs." }, { "id": "RCI043-O14", "field": "author_reported_limitations", "value": "algorithm comparison remains inconclusive; hyperparameters and reward coefficients were manually tuned and may not be optimal", "evidenceClass": "author-reported-limitation", "boundary": "Reported superiority must not be generalized beyond the disclosed tuning and configurations." } ], "disclosureAudit": [ { "id": "RCI043-D01", "field": "fixed_primary_artifacts", "status": "complete", "evidence": "HAL v3 and arXiv v1 file identities, page counts, byte counts and audit-date SHA-256 values are fixed.", "boundary": "The SSRN PDF bytes were not independently compared." }, { "id": "RCI043-D02", "field": "item_level_file_license", "status": "complete-with-restrictions", "evidence": "HAL v3 declares CC BY-NC-ND 4.0 for its PDF file.", "boundary": "No PDF redistribution, figure/table adaptation, extended-text reuse or commercial reuse of licensed expression." }, { "id": "RCI043-D03", "field": "public_code", "status": "not-found", "evidence": "No repository link appears on the fixed item pages or in the reviewed paper.", "boundary": "Absence on audited surfaces does not prove no private code exists." }, { "id": "RCI043-D04", "field": "public_raw_data_and_weights", "status": "not-found", "evidence": "No raw simulation trajectories, real-robot traces, trained policy, weights or result bundle is linked.", "boundary": "Author-reported plots cannot be independently recomputed." }, { "id": "RCI043-D05", "field": "exact_hardware_bom", "status": "not-disclosed", "evidence": "The physical platform is described at system level.", "boundary": "Motor, drive, encoder, cable, winch, motion-capture, compute and power part numbers are missing." }, { "id": "RCI043-D06", "field": "real_hardware_protocol", "status": "incomplete", "evidence": "One physical trajectory and qualitative setup are reported.", "boundary": "Repeat count, trial acceptance criteria, calibration, environmental conditions and failure log are absent." }, { "id": "RCI043-D07", "field": "real_hardware_metric_definition", "status": "incomplete", "evidence": "The paper reports a less-than-0.04 m position-error-norm trace.", "boundary": "No aggregation rule, confidence interval or downloadable values are supplied." }, { "id": "RCI043-D08", "field": "compute_identity", "status": "not-disclosed", "evidence": "Training duration is reported.", "boundary": "CPU, GPU, memory, software stack and energy measurement are absent." }, { "id": "RCI043-D09", "field": "software_environment", "status": "incomplete", "evidence": "OpenAI Gym is named.", "boundary": "No versioned environment, dependency lock, simulator version or reproducible container is supplied." }, { "id": "RCI043-D10", "field": "hardware_replication", "status": "not-disclosed", "evidence": "One real CDPR is reported.", "boundary": "No independent operators, multiple specimens or external replication are reported." }, { "id": "RCI043-D11", "field": "arbitrary_configuration_claim", "status": "not-demonstrated", "evidence": "Selected four- and eight-cable translational configurations are reported.", "boundary": "Actuator counts, geometry, mass, rotation and cable physics are not exhausted." }, { "id": "RCI043-D12", "field": "safety_validation", "status": "not-disclosed", "evidence": "A current-limit reward term is described.", "boundary": "Reward design is not functional-safety assessment or physical safety validation." }, { "id": "RCI043-D13", "field": "independent_rci_validation", "status": "not-performed", "evidence": "RCI inspected fixed paper artifacts and rights metadata.", "boundary": "RCI performed zero code, simulation, controller, actuator or robot runs." }, { "id": "RCI043-D14", "field": "component_or_procurement_conclusion", "status": "out-of-scope", "evidence": "The paper evaluates a control approach.", "boundary": "It cannot rank motors, drives, cables, computers or CDPR platforms or establish procurement suitability." } ], "method": "RCI fixed the official Inria/HAL v3 and arXiv v1 artifacts, recorded file identities and rights metadata, and converted fourteen paper-reported observations into independently worded fields with explicit boundaries. RCI did not reproduce figures or tables, execute code, run a simulator, operate a controller, test an actuator or access the physical robot.", "highestCompletedWorkLevel": "EV1 source verification", "limitations": [ "All control and performance results remain author-reported; RCI executed no code or hardware.", "The disclosed evidence covers selected translational point-mass configurations, not arbitrary CDPRs.", "The real-hardware result is one reported eight-motor trajectory without a complete repeatable protocol or raw trace.", "No public code, dependency lock, environment, weights, raw results or exact hardware BOM was confirmed.", "The 100 Hz setting is not a measured latency or real-time guarantee.", "HAL's CC BY-NC-ND 4.0 file license does not authorize commercial reuse or adaptation of the paper's expression." ], "license": { "rciOwnedData": "CC-BY-4.0", "rciOwnedCode": "MIT", "thirdPartyMaterials": "excluded; retain source-specific rights", "note": "This dataset contains RCI-authored factual abstraction, source metadata and original analysis. It does not redistribute or adapt the paper, figures or tables." }, "suggestedCitation": "Robot Component Index. “CDPR Actuator-Level Policy Evidence: 4-Cable Simulation to One 8-Motor Robot, Zero RCI Runs.” RCI 043, version 0.1.0, 2026-08-11. https://robotcomponentindex.com/research/cdpr-actuator-level-policy-evidence/" }