Who Earns the Safety? Intervention-Aware Quantum Predictive Control with Safety Attribution
2026-06-08 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors explain that simply using safety filters after a learned controller can hide whether the controller really understands safety or just relies on these filters. They introduce a new approach called IA-VQC-DPC that trains a quantum policy while limiting how much it depends on safety filters, and also measures how much safety comes from the policy itself versus the protective layers. Testing on building control simulations showed this method makes the quantum policy safer and more comfortable without using more energy, and better than a similar classical policy. Their safety-checking method can be used beyond just quantum controllers and buildings.
Safety filtersLearned controllersVariational Quantum Circuit (VQC)Intervention budgetControl Barrier Function (CBF)Policy trainingBOPTEST emulatorQuantum vs classical policiesSafety attributionRuntime guard
Authors
Yifan Wang
Abstract
Hard safety filters are increasingly placed downstream of learned controllers to guarantee constraint satisfaction at run time. Yet a filtered controller that never violates a constraint may still have learned nothing about safety: the filter can silently repair an incompetent upstream policy, so that post-filter success measures the filter, not the policy. We argue that safe policy learning should ask who earns the safety - the policy or its protective layers - and we make this question measurable. We introduce Intervention-Aware Variational Quantum Differentiable Predictive Control (IA-VQC-DPC), which (i) trains a compact variational quantum circuit (VQC) policy under a primal-dual intervention budget that penalizes reliance on a differentiable Control-Barrier-Function (CBF) projection, and (ii) is evaluated with a safety-attribution protocol that decomposes the executed-trajectory correction into a CBF term and a deployment runtime-guard term, and stress-tests the policy with guard-off evaluation. On closed-loop, high-fidelity BOPTEST building-control emulators (5 seeds, 60 episodes per method), intervention-aware training significantly lowers the quantum policy's raw pre-filter violation and total safety-layer reliance (both p < 10^-4) with no significant energy regression; at an equal approximately 400-parameter budget the quantum policy is significantly safer and more comfortable than a matched classical policy. Guard-off evaluation confirms the improvement is policy-level and exposes a valuable negative result: a learned differentiable energy head is only safe when paired with a distribution-aware runtime guard. The attribution protocol is general beyond quantum policies and buildings.