Open-Loop Consistency (OLC) in Control and Quantum Systems

• Open-Loop Consistency (OLC) is a property in control theory, quantum control, and string theory where precomputed (open-loop) solutions are shown to be equivalent to their feedback (closed-loop) counterparts under specific criteria.
• OLC bridges disparate frameworks, ensuring that methods like FBSDE and Riccati equation approaches yield consistent trajectories, equilibria, and amplitudes across differential games, quantum simulations, and string amplitude computations.
• OLC's practical implementation relies on rigorous conditions such as uniform convexity, precise discretization, and bounded error norms to guarantee that open-loop strategies faithfully reproduce closed-loop outcomes.

Open-Loop Consistency (OLC) is the phenomenon in control theory, differential games, quantum control, and string theory whereby the behaviors, solutions, or algebraic structures determined via open-loop (precomputed, non-feedback) protocols are, under specified conditions, faithfully reproduced—exactly or asymptotically—by a suitable closed-loop or feedback-realized implementation, and vice versa. OLC is central in domains where open-loop and closed-loop frameworks are structurally distinct but, under rigorous technical criteria, yield coinciding or reconcilable system trajectories, equilibria, or amplitudes.

1. Fundamental Definitions and Contexts

OLC arises in multiple mathematical and physical settings:

• Control Theory and Games: Open-loop controls are time-dependent functions (or stochastic processes) set in advance, whereas closed-loop (feedback) strategies adapt based on system state. OLC refers to the property that the set of trajectories or outcomes resulting from an open-loop protocol can be realized exactly (or up to asymptotic error) by a closed-loop mechanism constructed from the open-loop solution, and vice versa.
• Quantum Control: In quantum dynamics, continuous measurement imposes fundamental disturbance, often precluding direct feedback. The “Closed-Loop Designed Open-Loop Control” (CLOLC) protocol numerically simulates the feedback-designed law and applies open-loop pulses to the real system. OLC here denotes that, as the simulation is refined, the open-loop actuator's behavior converges to that of the ideal closed-loop system up to propagating the initial preparation error (Zhang et al., 2024).
• String Theory and Amplitudes: In the analysis of one-loop string amplitudes, OLC encodes the precise equivalence between mixed open+closed string amplitudes and sums over pure open string amplitudes, together with specific kernel factors, so that boundary terms and closed string insertions provide mutually consistent algebraic closures in monodromy relations (Stieberger, 2021).

2. OLC in Linear-Quadratic (LQ) Control and Differential Games

In stochastic LQ games, the system dynamics are linear in the state and controls, while the cost is quadratic. Three forms of equilibrium are relevant (Sun et al., 2016):

• Open-loop Nash equilibrium: Each player's control minimizes its cost, given the controls of others, with controls specified as stochastic processes.
• Closed-loop Nash equilibrium: Controls are feedback laws, typically affine in the current state.
• Closed-loop representation of open-loop equilibria: The open-loop Nash equilibrium admits a feedback (closed-loop) form if there exists a feedback law generating the open-loop trajectory for all valid initial states.

The critical results are as follows:

• Open-loop Nash equilibria are characterized by the solvability of a coupled FBSDE system; closed-loop Nash equilibria by a system of coupled symmetric Riccati ODEs.
• If the FBSDE admits a linear decoupling structure, the open-loop Nash equilibrium has a closed-loop representation via a non-symmetric Riccati ODE.
• In zero-sum games and LQ optimal control, the non-symmetric Riccati system collapses to the standard symmetric Riccati, and the closed-loop representation of open-loop solutions coincides with the closed-loop optimum—i.e., OLC holds exactly.
• The key assumptions for OLC are uniform convexity (and concavity in zero-sum), invertibility of the control cost matrix (or its generalization), and solvability and boundedness of the Riccati and BSDE structures (Sun et al., 2016).

3. OLC in Quantum CloLC Protocols

In quantum control, OLC manifests in the analysis of closed-loop designed open-loop protocols. The process involves:

• Designing feedback laws (often Lyapunov-based) that guarantee convergence to the target state in the ideal closed-loop setting.
• Classical simulation of the feedback law to generate a sequence of open-loop control pulses, discretized over $N$ grid points for a given time horizon.
• Applying these open-loop pulses to the real quantum system, which cannot be measured in real time without disturbance.

Theorems in (Zhang et al., 2024) establish:

• Asymptotic convergence: As the discretization $N$ increases, the only residual error in the final state is the unitary evolution of the initial preparation error:

$$\lim_{N \to \infty} e_N(T) = \Upsilon_{\rho_0}[T,0](\rho_0 - \sigma_0),$$

where $\Upsilon_{\rho_0}[T,0]$ is the propagator for the time-varying Hamiltonian of the targeted closed-loop system.

• Diminishing returns: Increasing the order of the numerical method beyond a threshold yields no further significant convergence improvement, as the dominant error term is $O(1/N)$ and is driven by the approximation error in the Hamiltonian.
• Uniform error norm bounds: The error between the closed-loop trajectory and open-loop realization is bounded explicitly in terms of discretization and initial state mismatch.

OLC thereby rigorously quantifies when and how open-loop application of simulated feedback laws can consistently track closed-loop quantum objectives (Zhang et al., 2024).

4. OLC in Discrete Sparse Optimal Control

In discrete-time sparse optimal control, the objective may be to generate inputs that drive the system to the target with minimum $\ell_1$-norm (sparsity-promoting), subject to dynamics and input constraints. OLC is realized as follows (Zhang et al., 2023):

• Open-loop solution: The optimal input sequence $u^*$ is obtained by minimizing a convex function over a finite horizon, with reachability constraints.
• Closed-loop realization: A dynamic linear compensator is constructed so that, for each basis initial state, the closed-loop application of this compensator reproduces $u^*$ exactly at every step, and the augmented closed-loop system is deadbeat stable.
• OLC property: In these systems, the open-loop and closed-loop solutions coincide (on the chosen basis), and the closed-loop system—constructed from the open-loop optimizer via a similarity transformation—guarantees both internal stability and trajectory reproduction.

Extensive algebraic construction ensures equivalence via the OLC basis and nilpotent closed-loop operators. Numerical studies confirm the analytic predictions (Zhang et al., 2023).

5. OLC in Linear-Quadratic Mean Field Games (MFG)

For large-population MFGs with LQ structure, OLC governs the connection between centralized open-loop/closed-loop Nash equilibria and decentralized, mean-field-optimal strategies (Liang et al., 18 Apr 2025):

• For finite $N$, open-loop NE (via FBSDEs) and closed-loop NE (via Riccati ODEs) differ and are structurally distinct.
• In the $N\to\infty$ limit, both open-loop and closed-loop centralized equilibria give rise to decentralized control laws characterized by the same limit Riccati equations.
• The asymptotically decentralized Nash equilibrium constructed from either approach yields the same state trajectories, feedback gains, and performance—realizing OLC in the mean-field limit.

A direct implication is that, for large populations, the choice between open-loop and closed-loop design is immaterial for optimal decentralized controller synthesis (Liang et al., 18 Apr 2025).

6. OLC and One-Loop Consistency in String Amplitudes

In string theory, OLC denotes the precise relationship between open and closed string amplitudes at one loop (Stieberger, 2021):

• One-loop amplitudes with both open and closed string insertions on the cylinder can be expressed as sums over purely open string amplitudes, with nontrivial kernel factors (the one-loop monodromy kernel).
• Analytic continuation and intersection theory for twisted cycles establish these relations, where a closed string insertion's contribution is identical to the boundary term in one-loop open string monodromy relations.
• OLC is the requirement that the algebraic closure of monodromy and dual “closed → open” mapping produce consistent physical amplitudes, ensuring the monodromy algebra is only closed by inclusion of the closed string channel.

In this context, OLC formalizes the necessity and structure of boundary terms as manifestations of closed string insertions and validates the analytic techniques connecting open and closed string sectors at one loop (Stieberger, 2021).

7. Implications, Limitations, and Domain-Specific Criteria

OLC is not universal without qualification. Its validity relies on structural properties of the underlying model and mathematical framework:

• Necessary conditions: Convexity/concavity of costs, solvability and boundedness of Riccati and BSDE structures, invertibility of key operators. In quantum control, accurate simulation to at least $O(1/N)$ in the Hamiltonian is essential; in MFGs, population size must diverge for OLC to manifest.
• Limitations: In finite-population games, open-loop and closed-loop domain representations are not in general consistent (the OLC property emerges only asymptotically); in classical control with strong state dependence or structural singularities, closed-loop realizability may fail; in quantum scenarios, measurement-induced disturbance can preclude direct feedback realization.
• Practical prescription: OLC provides explicit criteria for when open-loop simulation or optimization can be safely used to generate feedback-realized (or vice versa) controllers and elucidates the boundaries where simulation granularity, numerical method, or physical limitations dominate the residual discrepancy.

OLC thus serves as a unifying principle characterizing when disparate solution concepts in stochastic control, game theory, quantum dynamics, and field/string theory are reconcilable and computationally interchangeable, under provable criteria and structural constraints.