2209.07104
Identifiability Analysis of Noise Covariances for LTI Stochastic Systems with Unknown Inputs
He Kong, Salah Sukkarieh, Travis J. Arnold, Tianshi Chen, Biqiang Mu, Wei Xing Zheng
correctmedium confidence
- Category
- Not specified
- Journal tier
- Strong Field
- Processed
- Sep 28, 2025, 12:56 AM
- arXiv Links
- Abstract ↗PDF ↗
Audit review
The paper and the model derive the same decoupled measurement-difference setup (z_k, K with KH=0), the same vectorized linear system A[vec(Q); vec(R)] = vec(S), and the same rank-based identifiability conditions. The model’s arguments for the joint case (rank(KM)=p and rank(KCG)=g), the square case equivalence (rank(K)=p and rank(CAM)=p) implying B=D=0, and the Q-known and R-known cases match Proposition 2, Proposition 3 (and Corollary 1), and Proposition 4 in the paper. Where the paper is terse, the model fills in constructive proofs; there are no contradictions.
Referee report (LaTeX)
\textbf{Recommendation:} minor revisions
\textbf{Journal Tier:} strong field
\textbf{Justification:}
This paper delivers a rigorous, useful identifiability analysis for covariance estimation under unknown inputs using one-step measurement differences. The results are technically correct, well-motivated, and relevant to practice. Exposition is mostly clear, with some places where proofs are terse and notation could be streamlined. Modest improvements would further enhance clarity and accessibility.