Research Tasks

1 high-confidence edges lack evidence references. Edge IDs: e_v0_product_enables_capability (low_cost_parcel_sorting_robot_300k_rmb->affordable_small_warehouse_automation). Add evidenceIds with reviewed, high-confidence, non-vendor support or lower the edge confidence/review status if support is not available.

Frontier node IDs: industrial_area_scan_camera, machine_vision_lens_and_optics, controlled_machine_vision_lighting, vision_processing_compute, barcode_ocr_reading_software (+92 more). Expected node kinds: module, equipment, material, manufacturing_process, engineering_method, metric, bottleneck, evidence. Expand industrial_area_scan_camera first, adding the next layer where it affects maturity, cost, manufacturability, reliability, or deployment; add evidence where claims are medium/high confidence or leave explicit follow-up tasks for unsupported branches.

The scoped graph still marks decomposition frontiers such as industrial_area_scan_camera, machine_vision_lens_and_optics, controlled_machine_vision_lighting, vision_processing_compute, barcode_ocr_reading_software. Expand the next layer where it affects maturity, cost, manufacturability, or reliability.

ABB, Wayzim, Ambi Robotics, Berkshire Grey, and large MHE suppliers all appear relevant, but public data does not yet quantify shipped robotic parcel-infeed cells, revenue share, or installed base by supplier.

The robot graph now reaches copper, steel, aluminum, silicon, rare-earth, and elastomer feedstock chains, but it still lacks per-system quantities, supplier choices, price sensitivity, import dependence, and recycling assumptions.

The magnet branch now reaches NdPr and heavy rare-earth feedstocks, but it needs evidence for separation/refining, metal/alloy conversion, magnet sintering, coating, recycling, substitution, and price volatility.

The whole parcel-sorting robot has now been expanded across vision, end effector, reducer, servo drive, conveyor integration, and decision/control branches. The imported claims are unreviewed and need human review before being treated as settled product knowledge.

The suction/gripper branch is structurally decomposed, but pickup success, leakage, surface compatibility, blow-off timing, wear parts, and maintenance cost still need field or benchmark evidence.

The vision branch now has camera, optics, lighting, compute, and barcode/OCR software, but no-read rate, glare/occlusion robustness, latency, and parcel mix performance remain weakly evidenced.

The conveyor branch now includes sensors, speed tracking, spacing, PLC/WCS integration, chutes, and jam recovery, but installed cost, commissioning time, jam modes, and operator recovery time still need field evidence.

Queued from the graph research-root view for precision_reducer_gearbox. Follow docs/NODE_EXPANSION.md, research sources as needed, prepare a small candidate import batch, and add one deeper requires layer where it affects maturity, cost, manufacturability, reliability, or bottlenecks. Keep generated claims unreviewed until human review.

Vacuum cups, ejectors, valves, and pressure sensing are direct reliability and maintenance drivers, but current candidate data is mostly vendor product-line evidence without field wear rates or share data.

Nabtesco has a strong company-claimed global RV/robot-joint share, while public estimates for harmonic reducers and Chinese domestic RV reducers conflict by source and should not be mixed.

The reducer branch now separates wave generator, flexspline, circular spline, output bearing, and life testing, but precision machining, fatigue, lubrication interval, backlash growth, and cost evidence remain incomplete.

The first agent-assisted import decomposed the servo motor, but cost, supplier concentration, magnet supply, manufacturing yield, encoder alignment process, and robot-duty reliability are still weakly evidenced.

The industrial servo motor is currently a decomposition frontier, not a terminal node. Add stator, rotor, magnets, windings, encoder, housing, bearings, thermal design, motor-control interface, manufacturing processes, cost/supply constraints, and evidence.

The roadmap test products now continue past summary material-chain nodes into feedstocks, process equipment, manufacturing processes, and validation steps. Most of this second-pass structure is still supported only by low-confidence internal modeling and needs source-specific evidence before claims are promoted.

Six roadmap test products were added with recursive product-to-material/process decomposition. The claims are intentionally unreviewed and need human review before promotion.

The GLP-1 product graph reaches API synthesis, formulation, delivery devices, clinical package, and cold-chain capacity, but lacks reviewed evidence for peptide manufacturing routes, sterile fill-finish capacity, pen-device constraints, payer access, and long-term adherence.

The reusable space transport graph reaches booster reuse, upper-stage reuse, propulsion, materials, propellant, TPS, and operations, but the 1,000 USD/kg target needs reviewed cost, cadence, refurbishment, payload, reliability, and regulatory-throughput evidence.

The fusion graph reaches tokamak route, plasma control, magnets, plasma-facing components, tritium cycle, superconductor and tungsten/lithium material chains, but lacks plant-level evidence for tritium breeding, neutron damage, divertor lifetime, remote maintenance, and net electric availability.

The microreactor graph reaches core/fuel, TRISO fuel chain, heat transport, factory fabrication, power conversion, licensing, and high-temperature materials, but needs design-specific fuel qualification, security, transport, waste, refueling interval, and cost evidence.

The iPhone 4 graph reaches glass, semiconductor, battery, steel, and final assembly branches, but needs reviewed evidence for actual BOM suppliers, material mass, display/touch stack process, RF/enclosure issue history, and high-volume assembly yield.

The AK-47 graph is intentionally high-level and non-procedural, but needs reviewed historical evidence for production transitions, quality systems, material constraints, standardization, diffusion, and logistics impact without adding operational manufacturing instructions.