| """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 |
| from .systems import REGISTRY, build_plant, Plant |
| from .sensors import measure |
| from .metrics import settled_nrmse, episode_metrics |
| from .simulate import rk4_zoh_step |
|
|
|
|
| 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()) |
|
|
