Open-source libraries for designing controllers, identifying process models, and simulating plants, built on a verified Go BLAS/LAPACK. No cgo or Fortran toolchain needed.
Plant$ go get plantcontrol.org/v2/controlsys/v2
Each layer builds on the one below it: dense linear algebra, then state-space control, then identification and simulation tools.
A control-systems toolbox for linear time-invariant models, with state space as the core representation.
go get plantcontrol.org/v2/controlsys/v2github.com/PlantControl/controlsys →
PlantControl's fork of Gonum, focused on pure-Go BLAS and LAPACK, with the extra routines that control design needs.
go get plantcontrol.org/v1/gonumgithub.com/PlantControl/gonum →
Record what legacy numerical code does once, then replay those recordings against its modern replacement. Demonstrated with two SLICOT routines (MB01UD, SB03MD) ported to Go.
Native Go system identification: ARX, ARMAX, subspace (N4SID, MOESP, CVA), closed-loop and frequency-domain methods. Results keep a reproducible record of how each model was fitted.
A browser workbench for block diagrams, simulation, PID/LQG/H∞ design, and system identification. Ships as a single Go executable.
Control software runs real plants. Every library is held to the same rules.
No cgo and no Fortran runtime. A single go get cross-compiles anywhere Go runs.
Numerical results are checked against netlib LAPACK, SLICOT, and recorded oracles before they ship.
Code after the fork is AI-generated. Persistent tests, race checks, and reference comparisons gate every change.
Public libraries are MIT or BSD licensed. Upstream attribution is preserved where code descends from Gonum or netlib.
| Module | Repository |
|---|---|
plantcontrol.org/v1/gonum | gonum |
plantcontrol.org/v2/controlsys/v2 | controlsys |
plantcontrol.org/private/… | private repositories |
Private modules need Git access and go env -w GOPRIVATE=plantcontrol.org/private.
