stochasticIntegratorHarness

Shared test helpers for the stochastic-integrator equivalence tests.

Both equivalence suites (test_stochasticIntegratorsJulia.py and test_stochasticIntegratorsPaper.py) drive a native integrator on a small toy SDE with a prescribed sequence of Gaussian increments and compare the resulting trajectory against a committed reference. This module factors out the identical MuJoCo/StatefulSysModel scaffolding they share, so the per-suite files only supply the SDE (drift f and per-source diffusion columns gCols) and the reference case dict.

stochasticIntegratorHarness.buildPrescribedNoiseSim(case: dict, integratorClassName: str, f: Callable, gCols: List[Callable])[source]

Build a MuJoCo simulation that replays a reference case’s prescribed noise.

Parameters:

  • case – reference-case dict with keys dt, u0, n_states, n_steps, dW, dZ and (optionally) m. m defaults to n_states (the diagonal-noise convention used by the Julia suite).

  • integratorClassName – exact svIntegrators class name to instantiate.

  • f – drift f(t, x) -> length-n array.

  • gCols – list of m diffusion columns, each g_k(t, x) -> length-n array.

Returns:

(scSim, stateModel, integrator, stateLogger). Run to case[“tf”] and read stateLogger.x.