Concept note · the lab vocabulary
Adhesion
Holding onto a working surface that was never designed to be climbed.
What it is
Holding a machine against a working surface that was never designed to be climbed — compound curves, panel breaks, hardware in the path, and textures that fight a seal. The window bot solved the easy version: flat, smooth, continuous glass. The interesting problem starts where the glass ends.
In the lab
- Wash Bot →
- The reason the film’s shield-to-hood crossing bends a hinge instead of breaking physics: skirt vacuum held across panel breaks, validated in targeted multiphysics — 2 kPa on textured cladding under a 0.2 mL/cm² water film.
Our position
The seal is the product. A continuous dual-lip skirt holding roughly 2 kPa of vacuum — about 60 N of attachment on a 3.5 kg machine — matters more than any amount of drive torque, because everything else assumes it. The unlock is water-film adhesion: a thin film turns 80-grit-equivalent textured cladding into a sealable surface, which opens the entire SUV and crossover segment a dry skirt loses. Articulation serves adhesion, not the other way around — quadrants hinge ±15° so the skirt conforms to the curve instead of levering off it, and the machine lifts its own bristles to clear a door handle rather than break contact.
Honest limitations
Porous surfaces are a principled no: a soft-top convertible cannot hold vacuum, and no skirt geometry changes that — it is the documented exception that keeps the >99% fleet-coverage claim honest. The seal is proven in multiphysics simulation, not yet on a physical panel, and skirt wear over a service life — lip abrasion, contamination in the foam wiper — is engineering we have scoped but not modeled.
Related in the vocabulary
- Autonomy →
- Meets it in Wash Bot — machines that carry a job from start to finish without a hand on them.
The experiments are the proof
Every claim in this note is made concrete somewhere in the lab — the idea exists because the machines needed it.
The map of the lab →