Explicit-MPC Safety Shield for Neural Policies / README.md

Mechanism failed

Raw ⬇ ZIP

Explicit-MPC Safety Shield MVP

shield_experiment.py implements a deterministic toy wrapper for a disturbed 2D double integrator. It contains an affine LQR fallback, a polyhedral state check, the exact box-disturbance support term, robust one-step screening, and a smallest-grid correction toward the fallback.

Run:

/home/maxwelhelp/main/bin/python3 shield_experiment.py

The run writes results.json.