SCPN-Control

Neuro-symbolic control engine that compiles Stochastic Petri Nets into spiking neural network controllers. Five-tier gyrokinetic transport, 10 controller types, Grad-Shafranov equilibrium, JAX-differentiable.

130Python Modules
50Rust Modules
3,567Tests (100% cov)
10Controller Types
11.9 µsP50 Kernel Step
5GK Transport Tiers
20CI Jobs
33Equations Audited

Install

# Core (numpy, scipy, click)
pip install scpn-control

# With dashboard + JAX + machine learning
pip install "scpn-control[dashboard,jax,ml]"

# Everything
pip install "scpn-control[all]"

Core Pipeline

Petri Net
SPN structure
SNN Compiler
LIF + bitstream
Contracts
Pre/post checks
Controller
10 types
Plant Model
GS + transport

Key Capabilities

SPN → SNN Compilation
Stochastic Petri Nets compiled into LIF neurons with bitstream encoding. Contract-based pre/post-condition checking at runtime.
Grad-Shafranov Equilibrium
Fixed + free-boundary solver. Jacobi, SOR, multigrid. JAX-differentiable. Neural PCA+MLP surrogate (1000× speedup).
5-Tier Gyrokinetic Transport
Critical-gradient, QLKNN surrogate, native linear eigenvalue, native TGLF-equivalent (SAT0/SAT1/SAT2), nonlinear δf GK (5D Vlasov).
10 Controller Types
PID, MPC, NMPC, H∞, μ-synthesis, gain-scheduled, sliding-mode, fault-tolerant, SNN, PPO reinforcement learning.
Rust Acceleration
5 crates, 50 modules, 20,081 LOC. PyO3 bindings. Criterion-benchmarked: 11.9 µs P50 kernel step. Up to 2.5× speedup at scale.
JAX Autodiff
JAX-differentiable GS solver, transport, neural equilibrium. JIT compilation. GPU support. Gradient through complete equilibrium solve.

Explore