Q-DASC slashes HVAC comfort violations from 26% to 0.02% with quantum-classical hybrid safety layer

Variational quantum reinforcement learning offers compact policies for building energy control but suffers from a critical weakness: when the thermal model is locally incorrect, a policy that appears safe on the model can violate occupant comfort in the real building. To address this, Yifan Wang from the University of Michigan (preprint arXiv:2606.28834) proposes Q-DASC (Discrepancy-Attributed Safe Quantum Control). The system wraps a variational quantum circuit (VQC) policy with a certified classical safety layer that discovers misspecified operating regimes using false-discovery-rate control, repairs local thermal gains with shrinkage, projects the quantum schedule onto the repaired comfort-feasible set, and attributes residual violations to policy error, model error, or physical limits. Because the final certificate is produced by classical projection, comfort feasibility is invariant to finite-shot and depolarizing read-out noise—a critical advantage for near-term quantum hardware.

In experiments on real BOPTEST building emulators across three buildings, two localized misspecifications, and three seeds, Q-DASC reduced average comfort violation from 26.0% for the raw VQC controller and 55.3% for a model-trusting scheduler to 0.02%, matching a clairvoyant oracle with perfect model knowledge. Under NISQ read-out noise, violations remained at just 0.24%. A repair-aware VQC variant achieved 0.00% violation with lower projection intervention, while default Q-DASC maintained lower energy consumption and stronger observational-data behavior. The same safety wrapper transferred successfully to EnergyPlus heating and cooling benchmarks and to real hospital air-handling-unit data, establishing a safety-efficiency frontier for deploying quantum policies in physics-constrained control.