Repairability of Inexact Solvers in Recursive State Estimation with Machine Learning
📄 arXiv:2609.28425 · 📥 PDF · 2026-09-23 · quant-ph
Authors: Yanjun Ji [arXiv · scholar] , Dennis Willsch [arXiv · scholar] , Orkun Şensebat [arXiv · scholar] , Priyanka Arkalgud Ganeshamurthy [arXiv · scholar] , Zhi Pei [arXiv · scholar] , M. Sahnawaz Alam [arXiv · scholar] , Ivelina Stoyanova [arXiv · scholar] , Frank K. Wilhelm [arXiv · scholar] , Bo Zhao [arXiv · scholar] , Chao Wang [arXiv · scholar] , Kristel Michielsen [arXiv · scholar]
📄 Abstract
Recursive state estimation often executes approximate numerical solutions inside a feedback loop, where highly accurate local steps do not guarantee better overall results. For a fixed linear Kalman model, we characterize when a correction within a prescribed subspace and norm budget can meet a local admissibility tolerance, and how the defects actually executed affect the finite-horizon covariance response. Centering each defect on the exact gain for the implemented covariance separates current solve error from inherited gain drift. Expanding the exact residual-drift identity reveals opposing quartic contributions beyond the quadratic response: innovation-covariance inflation enters positively, while local-gain reoptimization enters subtractively. Under matched initialization, an absolute sixth-order remainder bound, uniform over bounded defect sequences at fixed horizon, gives sufficient conditions for quadratic under- or overprediction. Machine learning proposes bounded corrections, while a learner-independent residual certificate and verified fallback govern execution of classical and quantum candidates without changing the reference estimator. In a power-grid tolerance study, learned correction lowers the minimum conjugate-gradient iteration count for deployment without fallback relative to uncorrected solves under the same residual certificate. Gains reconstructed from a variational quantum linear solver and from an annealing-based binary encoding, with small-scale terminal measurements on superconducting hardware and sampling on a quantum annealer, are executed through the same interface. By linking local repairability to nonlinear error propagation, the framework evaluates approximate solvers and learned corrections through independent certification and finite-horizon response, providing a practical basis for studying hybrid quantum--classical computation.