SSC26 · The code

Repository: https://github.com/nscheuer/Generalized_ADCS

git clone https://github.com/nscheuer/Generalized_ADCS.git
cd Generalized_ADCS
pip install .

Run the poster’s code

Every snippet printed on the poster is executed verbatim by a verification script:

python papers/SSC26_poster/verify_snippets.py

It exits non-zero if any printed block stops working, so the poster and the library cannot drift apart.

Where each stage lives

Note

The adapter’s source sits under papers/Generalized_ACS/pipeline/ because it is that paper’s artifact, while installing and importing as ADCS.pipeline — so pip install generalized-adcs gives you the adapter either way. The legacy controllers (ADCS/controller/mtq_w_rw_LP.py and friends) remain as the validated baselines it is checked against.

Stage

Module

1 · Goal formulation

papers/Generalized_ACS/pipeline/goal_formulation/ — attitude_error.py, quat_set.py (reduced → full lift), omega_ref.py, world_vectors.py, conventions.py

2 · Control law

papers/Generalized_ACS/pipeline/control_law/ — law_interface.py (the contract), pd_law.py, sliding_mode_law.py

4 · Compensation

papers/Generalized_ACS/pipeline/compensation/ — gyroscopic.py, frame_rotation.py, disturbance_ff.py, damping_injection.py

5 · Allocation

papers/Generalized_ACS/pipeline/allocation/ — lp.py, qp.py, qpw.py, qpc.py, pseudoinverse.py, magnetic_cross.py, momentum.py

Orchestration

papers/Generalized_ACS/pipeline/pipeline_controller.py, papers/Generalized_ACS/pipeline/data.py

PipelineController subclasses the framework’s Controller, so it drops into ADCS.simulate and ADCS.mc.monte_carlo_runner.MonteCarloRunner without changes to either.

Design documents

The block interfaces are specified before they are implemented:

  • pipeline_spec.md — stage contracts and the LawInterface struct

  • goal_formulation_spec.md — goal → error-signal routing

  • allocation_spec.md — allocator formulations and momentum management

Tests

pytest testing/test_pipeline/

Includes test_pipeline_vs_lovera.py, which asserts the pipeline reproduces the published MTQ_Lovera controller to machine precision across six states including actuator saturation.

Reproducing published numbers

pyproject.toml declares compatible version ranges so the package installs into an existing environment. For deterministic campaign reruns use the exact pins instead:

pip install -r requirements-repro.txt
pip install -e . --no-deps

Also on the poster