Certified Design Regions under Parametric Plant Uncertainty with Application to Antithetic Integral Feedback
- 1Indian Institute of Technology Delhi
Abstract
Kinetic parameters in engineered biomolecular circuits are typically known only within intervals and are not precisely tunable; controller synthesis is therefore naturally posed over sets of gains rather than at a single nominal design. We formulate robust controller tuning under parametric plant uncertainty as a universally quantified set-inversion problem: given a plant-parameter box, compute the controller gains that stabilize every plant in the box. Robust stability is decided by Hurwitz sign conditions, which are bounded over the joint gain-parameter domain using validated range enclosures, yielding machine-checkable certificates of local exponential stability of the deterministic equilibrium. Numerical experiments show that the certified region depends strongly on the range enclosure. For an antithetic integral feedback (AIF) controller acting on a two-species gene-expression cascade, a tensor-product Bernstein enclosure certifies 83.2% of the design plane at bisection tolerance epsilon=0.05 with 324 terminal boxes, whereas four baseline enclosures certify between 50.9% and 58.9% on the same paving. For a fifth-order loop, the certified fraction increases from 34.4% to 73.8%, and the certified and excluded terminal boxes confine the robustly stabilizing area to a window of about 2.1 percentage points. Finally, for a refining variant of the algorithm in which the plant box is refined alongside the design box, the undecided area satisfies |U(epsilon)|=O(epsilon) under a non-degeneracy assumption on the stability boundary.
Related
- The p53-Mdm2 interaction through the lens of antithetic feedback
- Assessing measurement uncertainty at laboratory network scale and its impact on diagnostic performance in the absence of a gold standard: application to ELISA tests for Coxiella burnetii in ruminants
- WHEN MICROSCALE TRANSPORT CONTROLS MACROSCALE TISSUE FREEZING: A PARAMETRIC REGIME ANALYSIS
- Multi-parametric ex vivo magnetic resonance imaging for myelination in the developing mouse brain
- Multi-parametric NIR-II fluorescence perfusion imaging towards quantitative stroke evaluation
BibTeX
@misc{prakash2026certified,
title = {{Certified Design Regions under Parametric Plant Uncertainty with Application to Antithetic Integral Feedback}},
author = {R. Prakash and S. Sen},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.08.757569},
url = {https://www.biorxiv.org/content/10.64898/2026.10.08.757569v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Certified Design Regions under Parametric Plant Uncertainty with Application to Antithetic Integral Feedback
AU - R. Prakash
AU - S. Sen
PY - 2026
DA - 2026/10/08
PB - bioRxiv
DO - 10.64898/2026.10.08.757569
UR - https://www.biorxiv.org/content/10.64898/2026.10.08.757569v1
AB - Kinetic parameters in engineered biomolecular circuits are typically known only within intervals and are not precisely tunable; controller synthesis is therefore naturally posed over sets of gains rather than at a single nominal design. We formulate robust controller tuning under parametric plant uncertainty as a universally quantified set-inversion problem: given a plant-parameter box, compute the controller gains that stabilize every plant in the box. Robust stability is decided by Hurwitz sign conditions, which are bounded over the joint gain-parameter domain using validated range enclosures, yielding machine-checkable certificates of local exponential stability of the deterministic equilibrium. Numerical experiments show that the certified region depends strongly on the range enclosure. For an antithetic integral feedback (AIF) controller acting on a two-species gene-expression cascade, a tensor-product Bernstein enclosure certifies 83.2% of the design plane at bisection tolerance epsilon=0.05 with 324 terminal boxes, whereas four baseline enclosures certify between 50.9% and 58.9% on the same paving. For a fifth-order loop, the certified fraction increases from 34.4% to 73.8%, and the certified and excluded terminal boxes confine the robustly stabilizing area to a window of about 2.1 percentage points. Finally, for a refining variant of the algorithm in which the plant box is refined alongside the design box, the undecided area satisfies |U(epsilon)|=O(epsilon) under a non-degeneracy assumption on the stability boundary.
ER -