CFRP shell in the pagri silhouette, housing a 360° solid-state lidar turret and the head-compute cooling channel. The form is load-bearing — it is the sensor mount, not a cover over one.
Identity that
does work
The platform began as a Maratha warrior concept. Every cultural element survived into the engineering release because each one was given a job; nothing here is cladding.
Pagri crown
Sensor housingThe turban silhouette is a CFRP shell over a 360° lidar turret and the head-compute cooling duct. Remove the cultural form and you have to invent another mount.
Crescent aperture
Thermal vent and statusUpper-torso exhaust for the compute stack, doubling as the operator-facing autonomy indicator. Readable at twenty metres in daylight.
Patta bracer
Forearm tool moduleFold-out cutter, prybar, hex driver and micro-anchor in the geometry of the Maratha patta. Civilian tooling in a martial form.
Jali knit
Joint sheathingMughal-jali jacquard woven into the Dyneema knit at elbow and knee. It is the abrasion layer; the pattern is how we index panel orientation on the line.
Five surfaces,
one sole
Select a surface to see how the footpad and gait reconfigure. These are the five modes the platform is rated for.
Root and leaf litter. The planner favours trunk-adjacent footholds because they are load-tested by the tree.
The sole is the product
Most humanoids fail at the last 30 mm (the contact patch). ORION's footpad carries four force channels, retractable micro-cleats and a pad that spreads under load. The gait library is indexed against those three variables, which is why a new surface is a configuration exercise rather than a control rewrite.
Switching modes changes stride length, cadence, cleat extension and the acceptance threshold on contact confirmation. Nothing mechanical is swapped.
Five modes, one hardware setWhat it costs
to add a module
ORION Mk-I ships as a chassis with six interfaces. The controller re-identifies its own mass and inertia on power-up, so fitting a module is a bolt-on, not a recalibration.
Within envelope. All five terrain modes available.
Concept versus
what we build
The blueprint was theatrical, and beautiful for it. This is every place the engineering release departs from it, and why. No number here is a marketing number.
| Item | Concept blueprint | ORION Mk-I | Why it moved |
|---|---|---|---|
| Stature | 1.95 m | 1.78 m | Mass scales with the cube of stature. 1.78 m keeps hip torque inside available SEA curves. |
| Mass, dry | 210 kg | 96 kg | A 210 kg biped needs actuators that do not exist at this price. Armour became structure instead of cladding. |
| Runtime | 18 h | 4.5 h active | Solid-state Li-metal at present density gives 3.6 kWh in the torso envelope. 18 h was never in the cells. |
| Actuation | Unspecified | SEA + QDD hybrid | Compliance at the contact joints, stiffness at the wrist. One scheme for both would compromise each. |
| Terrain modes | 5, mechanical | 5, software-selected | Same five surfaces, but the sole reconfigures under control rather than by tool change. |
| Patta | Bladed weapon | Utility tool module | The platform is a civilian field-services humanoid. The form stayed; the function changed. |
What we
have not solved
Published because a specification without its open items is a brochure.
Pack swap time
The target is a two-minute hot swap. The current prototype takes fourteen minutes because the dorsal cowl is fastened, not latched.
MPC latency on deformable ground
200 Hz whole-body control with lidar in the loop is feasible on Thor-class compute on paper. Worst-case latency on wet clay is the open variable.
Cultural representation review
The pagri silhouette and crescent are deliberate. Before any public unveil the design goes to cultural advisors: the references must read as honouring, not appropriating, including in export markets.
Bracer as identity vs. modularity
If most customers swap the patta bracer for a generic module, the identity attenuates. Open question whether the bracer geometry becomes the standard tool form factor instead.