Validation Methodology

This section describes how GoMyRobotOS validates its own claims: reproducible boot, functional runtime validation, temporal isolation, spatial isolation, interference, fault injection, and recovery timing.

First rule: what validation is and is not

Warning

No certification claims exist at M0 GoMyRobotOS validation is an engineering research methodology. The documentation does not claim:

  • formal WCET

  • certification

  • safety certification

  • complete freedom from interference

…unless and until actual evidence exists from the experiments described here. Every page in this section carries the same rule as the rest of the project: Status labels only, no invented results.

The declared validation areas

Area

Page

What it validates

reproducible boot

M1

same contract → same boot, machine to machine

functional runtime validation

M1

partition boots, runs, communicates

temporal isolation

Temporal isolation

budget respected under load

spatial isolation

Spatial isolation

declared memory/IRQ/DMA/device boundaries hold

interference

Interference

cross-partition effects, multi-channel

fault injection

Fault injection

contract recovery semantics under fault

recovery timing

Recovery

T_detect / T_contain / T_recover / T_resume

How the results feed the rest of the ecosystem

validation experiments (this section)
        │
        ▼
test / evidence artifacts
        │
        ├──→ GoMyRobotVerify   (result processing, coverage, traceability)
        └──→ GoMyRobotAssure   (evidence graph, assurance argument)

GoMyRobotOS produces the artifacts; the external products interpret them (ADR-0012). That is why this section is careful with two words: “validated” (a runnable experiment existed and its result is recorded) versus “demonstrated” (the acceptance gate of a milestone passed).

Ladder of where validation runs

host CI (x86-64)
   → QEMU (reproducible environment)
   → NG-ULTRA hardware (flight reference)
   → HPSC (research)

Measurements move up this ladder only when a lower-level measurement calls for hardware-specific evidence; the ladder is the same one GoMyRobotBench uses for physical experiments.