# Эксперимент: HOCBF Safety Shield for Neural Policies (#949)

{
  "worked": true,
  "confidence": 8,
  "verdict": "Built an analytic 2-D double-integrator HOCBF safety shield with actuator-bounded Euclidean projection and reproducible numerical experiments. The derivative identities were confirmed with maximum absolute errors of 2.7e-6 for h-dot and 1.4e-5 for psi1-dot; required normal acceleration scaled exactly linearly with k2, with observed slope -0.170955 versus predicted -psi1=-0.170955. Shield activation began at the predicted boundary transition, and shielded rollouts maintained positive minimum h while unshielded rollouts collided. This demonstrates the mechanism in the toy setting, but not improved learned-policy reward.",
  "metrics": {
    "baseline": "Unshielded fixed policy: minimum h approximately -0.816 to -0.820 across tested scales and timesteps.",
    "idea": "HOCBF shield: minimum h approximately 0.0038 to 0.0056, activation rate 0.884–0.968, infeasible rate 0.0; derivative errors 2.7e-6 and 1.4e-5; observed k2 slope -0.170955 versus predicted -0.170955."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 hocbf_experiment.py",
  "files": [
    "hocbf_experiment.py",
    "results.json",
    "run_output.txt"
  ],
  "limitations": "Only a known-model 2-D point-mass system and one fixed nominal policy were tested. No PPO/SAC training, differentiable QP backpropagation, model uncertainty, multiple obstacles, disturbances, or repeated statistical trials were evaluated; discrete-time safety results are not a formal continuous-time guarantee."
}