Rust-native robotics simulation

Design. Test.
Decide.
Before Metal.

Design robots, run reproducible experiments, and see every model's evidence status, validity limits, and unqualified gaps.

  • 01 Reproducibility specified
  • 02 Rust-owned core
  • 03 Evidence-aware
Example BeforeMetal experiment definition for an outdoor rover weather baseline.
EXPERIMENT / 001rover.weather-baseline
Preflight
RBRobotoutdoor_rover.v1defined
WDWorldwet_slope_gustdefined
CTControllertraction_control.v2linked
FDFidelityengineeringreview
33Normative sections
683Named requirements
27Acceptance gates
5Documented feedback loops
Source of truthrequirements.mdSite generated at build time

The decision loop

From model to evidence.

BeforeMetal treats simulation as a traceable engineering argument—not a black box that emits an authoritative-looking animation.

  1. 01

    Describe

    Robot, world, controller, mission, and design variables.

  2. 02

    Compile

    Capabilities, timing, ports, fidelity, and unsupported features.

  3. 03

    Run

    A reproducible experiment with explicit models and seeds.

  4. 04

    Compare

    Hardware, controllers, weather, faults, and parameter sweeps.

  5. 05

    Validate

    Compare predictions with available evidence and disclose gaps and validity limits.

Evidence contract

Know what produced the result.

BeforeMetal is designed to preserve the configuration, assumptions, runtime record, and available comparison evidence—or its explicit absence—behind each engineering decision.

Inspect the evidence requirements
  1. DATA-001

    What ran

    A manifest records the project, compiled scenario, controller, software, seeds, execution mode, fidelity, and termination state.

    Run manifest
  2. VAL-001

    What it assumed

    A fidelity card lists active models, property dispositions, approximations, validity warnings, and available evidence or its explicit absence.

    Fidelity card
  3. SYS-011

    What happened

    Simulated time, wall time, real-time factor, solver diagnostics, warnings, and termination reason stay visible.

    Run diagnostics
  4. DATA-013

    What changed

    A comparison report identifies configuration, software, seed, fidelity, and selected metric differences between runs.

    Difference report

Portable MVP

A rover. Weather approximations. A bounded promise.

The first useful product is a local, Rust-only simulator for an outdoor rigid-link rover with a basic viewer and a complete headless workflow that requires no physical robot or measurement hardware.

Read the MVP definition

In the portable baseline

  • Rigid mechanics, gravity, terrain, wind, and gusts
  • Approximate rain, wetness, and surface effects
  • Motors, gearing, batteries, sensors, and faults
  • Rust controllers, sweeps, results, and fidelity cards
  • Digital benchmark qualification without a physical robot

Not required for MVP

  • Python, ROS 2, MuJoCo, Gazebo, or FMI
  • Production CFD, FEA, SPICE, or CAD
  • Photorealism or a discrete GPU
  • A claim of universal physical accuracy or certification

Definition before implementation

See exactly what BeforeMetal is committing to.