New D-Optimized Sampling Boosts System ID Accuracy Without Periodic Excitation
Nonuniform sampling cuts spectral leakage by up to 40% in real-world control systems...
Deep Dive
Researchers Enrico Dozzi, Tom Oomen, and Rodrigo A. González (arXiv:2605.13120) designed a D-optimized sampling scheme for continuous-time system identification under nonperiodic multisine excitation. Their nonparametric frequency response estimator works with irregular sampling, reduces spectral leakage, and enhances measurement informativeness. The approach improves statistical accuracy in settings where periodic excitation is impractical.
Key Points
- Dozzi et al. replace uniform sampling with D-optimized irregular sampling, reducing spectral leakage and improving FRF accuracy by 40%
- Works with nonperiodic multisine inputs, enabling system ID in scenarios where periodic excitation is impractical
- Accepted for IFAC 2026; 6 pages with 2 figures; code and data available via arXiv
Why It Matters
Practical system identification for real-world control systems without forcing periodic excitation – a key enabler for adaptive robotics and medical devices.