Trust is earned.
Trustworthiness is measured.

A physics-grounded framework for evaluating autonomous system trustworthiness.

Kinetic Energy
The energy a vehicle carries. The focal variable in safety — the agent of harm in every crash and every outcome.
Tolerance
What the surrounding environment can manage when things go wrong — through conflict control, containment, and absorption.
Margin
The gap between energy and tolerance. A trustworthy system holds it with discipline — neither too thin nor unnecessarily wide.
Backing into a spot
Read the framework → See it in action
Continue
Built for
AV Developers
Translate your stack's complexity into the trustworthiness riders care about.
Fleet Operators
See where margins run thin across routes, conditions, and AV providers.
Autonomy Platforms
A shared trustworthiness signal across simulation, deployment, and domain.
Product
Static Foundation
Tolerance File
Per-road-segment nominal tolerance thresholds that plug into existing map and planning stacks.
Dynamic Layer
Tolerance Model
Adjusts nominal tolerance based on current conditions — weather, road users, and temporary hazards — in simulation or on the road.
Use it to
01
Incident Context
Show whether a collision stayed within tolerance — or crossed it.
02
Test Prioritization
Point testing at the routes and conditions where margin is thinnest.
03
ODD Expansion
Evaluate whether margin holds in faster, denser, or less forgiving domains.
04
Margin Benchmarks
Compare margin across routes, conditions, and operators on common ground.
05
Shared Signal
Give every team — engineering, safety, product, policy — the same number.
Coming Soon

Beyond vehicles. Physical AI needs the same lens.

Robotics, drones, industrial automation — any system governing kinetic energy in the real world.

→