| name | eigenstate-reduction-quantum-smcp |
| description | Tractable Infinite-Horizon Stochastic Model Predictive Control for Quantum Filtering via Eigenstate Reduction. Uses almost-sure eigenstate reduction of quantum trajectories to collapse infinite-horizon stochastic objective to closed-form fidelity term. Eliminates per-horizon Monte Carlo sampling while retaining stochastic dynamics. Use when designing quantum SMPC controllers, quantum filtering-based control, mean-square stability analysis, or tractable stochastic optimal control for quantum systems.
|
| metadata | {"arxiv_id":"2511.05916","published":"2025-11-08","authors":"Yunyan Lee, Ian R. Petersen, Daoyi Dong","tags":["quantum-control","MPC","stochastic-control","quantum-filtering","eigenstate-reduction","optimal-control","systems-engineering"]} |
Eigenstate-Reduction Quantum SMPC
Overview
Stochastic Model Predictive Control (SMPC) for quantum systems traditionally requires per-horizon Monte Carlo
scenario sampling, which is computationally prohibitive. This methodology uses the almost-sure eigenstate reduction
of quantum trajectories under continuous measurement to collapse the infinite-horizon stochastic objective to a
closed-form fidelity term computable from the one-step averaged state.
Core Mathematical Insight
Eigenstate Reduction Theorem
Under continuous-time measurement, quantum trajectories almost surely converge to eigenstates of the measurement
operator. This means the infinite-horizon expected cost:
J = E[∑ γ^k · cost(ρ_k, u_k)]
collapses to:
J = F(ρ̄_1) (closed-form fidelity from one-step averaged state)
where ρ̄_1 is the one-step averaged (expected) state after applying the control.
Computational Reduction
| Approach | Complexity | Scalability |
|---|
| Sampling-based SMPC | O(N_scenarios × N_horizon) | Poor for long horizons |
| Eigenstate-reduction SMPC | O(1) terminal evaluation | Linear in system dimension |
Algorithm
Step 1: Quantum Filter Propagation
ρ̄_{k+1} = 𝔼[ρ_{k+1} | ρ_k, u_k] (deterministic)
Propagate the averaged density matrix deterministically — no sampling needed.
Step 2: Terminal Fidelity Evaluation
F_term = Tr(ρ_target · ρ̄_N)
Evaluate fidelity against target state at the prediction horizon endpoint.
Step 3: Optimization
min_u -F_term(ρ̄_N(u))
s.t. physical constraints on u (amplitude, bandwidth, etc.)
Solve the deterministic optimization problem.
Step 4: Receding Horizon Implementation
Apply first control step, re-measure, re-optimize.
Stability Guarantees
- Equivalence: The reduced objective is equivalent to the full stochastic objective in the infinite-horizon limit
- Mean-square stability: Closed-loop system is mean-square stable under the receding horizon policy
- Validated on multi-level systems and Ising-type coupled systems
Application Workflow
- Define system: Hamiltonian H₀, control Hamiltonians H_c, measurement operators M
- Set target: Desired quantum state or gate operation
- Configure filter: Quantum stochastic differential equation for averaged state propagation
- Terminal cost: Fidelity with respect to target state
- Solve: Standard optimization (gradient-based or derivative-free)
- Validate: Compare with sampling-based SMPC for correctness
Key Advantages
- No Monte Carlo sampling: Eliminates the dominant computational bottleneck
- Closed-form terminal: Fidelity computable analytically from one-step averaged state
- Mean-square stability: Theoretical guarantees preserved
- Scalability: Linear scaling vs exponential for sampling-based approaches
- Generality: Applicable to finite-dimensional quantum systems under continuous measurement
Pitfalls
- Eigenstate reduction assumption: Requires continuous measurement to drive trajectories to eigenstates. Intermittent or weak measurement may not satisfy the convergence condition.
- Finite-dimensional systems: Theory established for finite-dimensional Hilbert spaces. Extension to infinite dimensions requires additional analysis.
- Measurement backaction: The approach assumes the measurement operator structure is known and correctly modeled.
- Horizon truncation: While theoretically infinite-horizon, practical implementation uses finite N — verify convergence.
Related Skills
vf-qctrl-llm-quantum-control — LLM-driven quantum control (alternative approach)
quantum-control-pulse-software — Pulse-level control software framework
model-predictive-quantum-control — General MPC for quantum systems
discounted-mpc-control — Discounted MPC stability analysis (classical analogue)
Activation: stochastic model predictive control, quantum SMPC, quantum filtering, eigenstate reduction, quantum trajectory, continuous measurement, mean-square stability, quantum optimal control, receding horizon quantum control