Abstract
We propose a quantum-assisted framework for solving constrained finite-horizon nonlinear optimal control problems using a barrier Sequential Quadratic Programming (SQP) approach. Within this framework, a quantum subroutine is incorporated to efficiently solve the Schur complement step using block-encoding and Quantum Singular Value Transformation (QSVT) techniques. We formally analyze the time complexity and convergence behavior under the cumulative effect of quantum errors, establishing local input-to-state stability and convergence to a neighborhood of the stationary point, with explicit error bounds in terms of the barrier parameter and quantum solver accuracy. The proposed framework enables computational complexity to scale polylogarithmically with the system dimension demonstrating the potential of quantum algorithms to enhance classical optimization routines in nonlinear control applications.
Introduction
Nonlinear Optimal Control Problems (OCPs) are essential for modern engineering, yet they suffer from the "curse of dimensionality." Standard solvers like Sequential Quadratic Programming (SQP) iteratively solve Karush-Kuhn-Tucker (KKT) linear systems, which classically scale cubically ($O(n^3)$) with the number of states and controls. This makes real-time control of large-scale systems—such as power grids or complex biological models—computationally prohibitive. This paper proposes a hybrid quantum-classical framework that offloads the most expensive linear algebra steps to a quantum subroutine, achieving exponential speedup in system dimension while maintaining classical stability guarantees.
Methods: Hybrid Quantum-Classical SQP
The proposed framework utilizes a classical "outer loop" for high-level logic and a quantum "inner loop" for solving linear systems.
Barrier SQP and Schur Complement
The algorithm uses a barrier method to handle inequality constraints by incorporating them into a smooth augmented objective function. To optimize the quantum step, the authors apply a Schur complement reduction, which reduces the size of the KKT system that the quantum computer must solve.
Quantum Schur Step
The core innovation is the use of Quantum Singular Value Transformation (QSVT) to perform matrix inversion. This involves:
- Block-Encoding: Embedding the problem's matrices into a larger quantum unitary operator.
- QSVT Primitive: Applying a polynomial transformation to the singular values of the block-encoded matrix to compute the solution.
To ensure reliability, the authors provide a formal Input-to-State Stability (ISS) analysis. This proves that despite the noise and approximation errors inherent in quantum computing, the algorithm will reliably converge to a neighborhood of the true optimal solution.
Results and Validation
The framework was validated using a simulation of the Perelson HIV-1 model, a biological control problem that models interactions between uninfected T cells, infected cells, and free virus particles. The goal is to determine optimal dosages of Reverse Transcriptase Inhibitors (RTIs) and Protease Inhibitors (PIs) to minimize viral load.
As shown in the simulation results, the hybrid quantum-classical solver produced trajectories nearly identical to a pure classical SQP solver. This confirms that the "inexact" nature of the quantum step does not degrade control quality in practice. Theoretically, the quantum subroutine scales polylogarithmically ($O(\text{polylog}(n))$) with the system dimension, offering a significant advantage over classical solvers for high-dimensional systems.
Discussion and Limitations
While the polylogarithmic scaling is a major breakthrough, the runtime remains sensitive to the condition number ($\kappa$) of the matrices. Ill-conditioned problems can diminish the quantum speedup. Furthermore, implementing these algorithms on current Noisy Intermediate-Scale Quantum (NISQ) hardware remains a challenge due to the high gate fidelities and coherence times required for QSVT. Future research will focus on "warm-starting" techniques to further accelerate the quantum solver across SQP iterations.
Conclusion
This work demonstrates that quantum-assisted SQP can successfully bridge the gap between quantum algorithms and classical nonlinear control. By providing explicit mathematical bounds on convergence and demonstrating practical accuracy in biological simulations, the authors establish a robust foundation for applying quantum computing to large-scale optimization challenges.
Summary updated