Datasets:
File size: 1,397 Bytes
95434b3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 | """Standalone plant, sensor and metric definitions of the HyperKKL observer benchmark.
from hyperkkl_bench import build_plant, load_meta
meta = load_meta("vdp") # metadata/vdp.json of the dataset
true_plant = build_plant(meta["system_config"], perturb=0.0) # C1 / C3-partial data
real_plant = build_plant(meta["system_config"], perturb=0.3) # C3-mismatch30 / C3-realistic data
model_exact = build_plant(meta["system_config"]) # modeller, knowledge = full
model_part = build_plant(meta["system_config"], knowledge="partial") # f_known (UNMODELLED term missing)
Requires torch, numpy (scipy for the colored_t sensor model).
"""
import json
from pathlib import Path
from . import systems_extra, systems_bounded # noqa: F401 (register battery, network, cartpole_b)
from .systems import REGISTRY, build_plant, Plant # noqa: F401
from .sensors import measure # noqa: F401
from .metrics import settled_nrmse, episode_metrics # noqa: F401
from .simulate import rk4_zoh_step # noqa: F401
def load_meta(plant: str, root=None) -> dict:
"""Read metadata/<plant>.json from a local copy of the dataset (default: ../../metadata)."""
root = Path(root) if root else Path(__file__).resolve().parents[2]
return json.loads((root / "metadata" / f"{plant}.json").read_text())
|