SCPN-Phase-Orchestrator

Domain-agnostic coherence control compiler built on Kuramoto/UPDE phase dynamics. One mathematical engine — 36 domains. From plasma physics to cardiology, power grids to neuroscience.

PyPI Package Live Demo 12 Engines 36 Domains

760Python Modules

119Rust Modules

1,070Test Files

603Rust Tests

99%Test Coverage

12UPDE Engines

36Domainpacks

19Monitors

Install

# Core framework (4 dependencies: numpy, scipy, pyyaml, click)
pip install scpn-phase-orchestrator

# With Rust acceleration + differentiable backend
pip install scpn-phase-orchestrator[rust,nn]

# Everything (quantum, plasma, fusion, JAX, gRPC, FPGA)
pip install scpn-phase-orchestrator[full]

# Standalone Rust kernel (stable-ABI wheels, five platforms)
pip install spo-kernel

Core Pipeline

Domain Binder

YAML spec

→

P/I/S Extractors

Physical · Informational · Symbolic

→

UPDE Engine

12 variants

→

Supervisor

MPC + Regime FSM

→

Actuation

Control output

4 Universal Control Knobs

K — Coupling Strength

Inter-oscillator coupling magnitude. Controls how strongly oscillators influence each other. Mapped to domain-specific parameters via binding spec.

α — Phase Lag

Sakaguchi-Kuramoto phase frustration parameter. Shifts the preferred phase relationship between coupled oscillators.

ζ — Driver Strength

External forcing amplitude across P/I/S channels. Controls how strongly the system is driven by reference signals.

Ψ — Reference Phase

Target phase for the driver. Sets the desired synchronisation state. The system converges toward this reference.

Key Capabilities

Cross-Domain Universality

36 domainpacks from plasma physics to cardiology, power grids to EEG, financial markets to swarm robotics. Same engine, same 4 knobs, different binding specs.

Closed-Loop Control

Built-in MPC supervisor, regime FSM (NOMINAL/DEGRADED/CRITICAL), Petri net state machine, Hebbian plasticity. Not just simulation — active control.

Differentiable Backend

JAX-based nn/ module. Fully differentiable KuramotoLayer, Stuart-Landau, UDE, inverse coupling inference. JIT + vmap + GPU.

Rust Acceleration

119 Rust kernel files via PyO3 FFI. 2.4×–96× speedup. Pure Python fallback ensures the package works without native extension.

Production Stack

gRPC server, Docker/Helm, Prometheus metrics, Redis state, SHA256 audit trail, STL runtime monitoring, OpenTelemetry export.

Hardware Targets

FPGA (Zynq-7020, sub-15µs latency), GPU (JAX/CUDA), WebAssembly (browser demo), bare metal deployment.

Validated on Real Grid Data

Grid Sentinel Replay →

Interactive replay of the live modal sentinel over real ISO-NE PMU captures at a sealed, hash-verified operating point. Review-only alarms; the claim boundary is stated on the page.

Modal Head-to-Head →

The domain-specific modal envelope-growth detector against four generic early-warning detectors on the identical matched-false-alarm protocol.

The Honest Test →

Why a matched false-alarm budget with label-permutation significance is a stricter bar than the retrospective per-record test, shown step by step.

Preprint: Generic early-warning signals at chance

Five modalities, one matched-false-alarm protocol, permutation-tested nulls. Zenodo doi:10.5281/zenodo.22113062 (CC-BY). Under review at Journal of the Royal Society Interface.

Preprint: Modal-growth detector + eigenvalue regime map

The grid positive result, its deployment bounds, and external validation against small-signal eigenvalues. Zenodo doi:10.5281/zenodo.22113116 (CC-BY). Under review at IEEE Transactions on Power Systems.

Sealed evidence records

Every alarm and non-detection binds to content-addressed, byte-reproducible artefacts under examples/real_data/; committed integrity tests recompute each sealed hash from the payload alone.