The control loop runs in the cloud. The plant data and control inputs must stay secret. Quantum keys make explicit MPC encrypted and lightweight.
Mi, Ren, Wang, Quevedo, and Shi (arXiv:2603.22653) design quantum-encrypted explicit model predictive control for constrained linear systems in a cloud-based setup. The standard problem: solving MPC online is computationally expensive, so you offload it to a cloud server. But the server sees your plant state and control inputs — proprietary process data you don't want exposed. Classical encryption works but adds computational overhead that can violate real-time constraints.
Explicit MPC resolves the online computation problem by solving the optimization offline, producing a piecewise-affine control law — a lookup table indexed by state regions. The online step reduces to evaluating which region contains the current state and computing a linear function. This is fast but exposes the structure: the region boundaries and affine gains encode the system dynamics, constraints, and cost function.
The quantum approach uses Bell pairs for key generation via quantum key distribution. The encryption is one-time-pad — information-theoretically secure, not just computationally secure. The encrypted evaluation recovers the plaintext control action exactly, with no approximation error from the encryption layer.
The result is reduced online complexity compared to classically encrypted MPC. Quantum key distribution handles the key management problem that plagues classical one-time-pad schemes (you need as many key bits as data bits, and they can't be reused). The quantum channel generates keys at the rate needed by the control loop, synchronized to the sampling period. The control problem generates the key consumption schedule. The security is physical rather than computational — no amount of classical computation breaks it.