Roboply
Robotic hands for the composites factory — trim, drill, handle, assemble.
Vision-guided autonomous robots for ply handling, tool loading and part movement — plus robotic trimming, drilling and bonding — driven by Isaac perception and coordinated with the rest of the Curelum loop.
Why this step condemns parts
Around layup and cure sits a chain of dangerous, precise, craft-bound manual work: ply handling and kitting, tool loading, part movement, and post-cure trimming, drilling and bonding. Carbon dust is hazardous, tolerances are tight on flight-critical parts, and scarce cleared technicians spend hours on repetitive tasks. Fixed automation is brittle and cannot adapt to part variation or the positional data the rest of the factory already senses.
Robotic hands for the composites factory — trim, drill, handle, assemble.— Roboply, part of the Curelum autonomy platform
A closed perceive → decide → act loop
No open-loop hoping. Roboply senses, reasons against the qualified envelope, acts within bounds, and logs every action.
Perceive the part
Isaac-based perception locates the part, its features and its as-built geometry from the shared Curelum context.
Plan the action
Grasp, trim, drill or bond paths are planned against the part’s real geometry and defect map — not a fixed program.
Execute to tolerance
Robots handle, kit, load, move, trim, drill and bond to flight-critical tolerance, adapting to variation.
Coordinate the cell
Actions are sequenced with Plyoq, Curevo and Yieldex so robots act on the same context as the rest of the line.
What Roboply does on the line
Vision-guided handling
Ply handling, kitting, tool loading and part movement, autonomously.
Robotic trim / drill / bond
Post-cure trimming, drilling and bonding to ±0.1mm tolerance.
Context-aware actions
Uses Plyoq placement and Voidra defect maps to adapt in real time.
Safer operations
Removes technicians from hazardous carbon-dust and repetitive tasks.
Isaac + Jetson
GPU perception and grasping on the factory edge.
Action traceability
Every robotic action logged to the part record.
Instrumented, not guessed
Every Roboply readout maps to a real measurement on the floor — tow geometry, cure trajectory, C-scan integrity, or predicted strength. The interface is precision test equipment: numbers you can certify, not a dashboard of vanity charts.
- Grounded, auditable outputs on every action
- Bounded control envelopes from the Strandix twin
- Human-in-the-loop on flight-critical decisions
- Immutable audit trail for AS9100 / NADCAP
Built for the composites floor
Roboply deploys on the factory edge and integrates with the systems you already run, under your security and export controls.
| Tasks | Handling, kitting, tool load, movement, trim, drill, bond |
|---|---|
| Perception | Isaac-based vision & grasping |
| Tolerance | ±0.1mm on flight-critical features |
| Adaptation | Part-variation aware, no fixed re-tooling |
| Coordination | Shared context with Plyoq, Voidra, Curevo, Yieldex |
| Compute | Jetson factory-edge, Isaac perception |
Fits your stack, under your controls
AFP / ATL controllers
Autoclave / OOA
Ultrasonic & thermographic NDT
MES / ERP
OPC-UA / OT
Historian & PLM
Trust before autonomy
Roboply runs in shadow, then assist, then bounded autonomy — grounded, auditable, and human-supervised at every airworthiness-critical checkpoint.
Stronger together across the loop
Questions engineers ask
Do we need new robots?
Roboply drives industrial robot arms with Isaac-based perception; it can work with common robot platforms rather than requiring bespoke hardware.
How is it safer than fixed automation?
It adapts to part variation using shared perception, so it handles real geometry rather than assuming a fixed program, and it removes people from hazardous dust and repetitive tasks.
Does it coordinate with the rest of Curelum?
Yes — it consumes placement and defect context so robots act on the same part timeline as layup, cure and NDT.
Put Roboply on your line
Start with one cell, a hard success metric, and a paid pilot. Prove ROI, then expand across the loop.