Structured nonlinear inversion of systems with memory: a representation-aware constructive principle across operator, structural, and geometric views
The inversion of nonlinear systems with memory is challenging because inverse operators can be difficult to construct and implement efficiently. Volterra-series methods provide a flexible functional framework, but dense high-order kernel realizations may suffer from combinatorial complexity. Physical approaches, including circuit-based inversion and Wave Digital Filter (WDF) implementations, can instead exploit structural locality. This Hypothesis and Theory article argues that recursive Volterra inversion, structural circuit inversion, and geometric approaches based on Lie theory, while not equivalent, share a constructive principle: inverse synthesis through local, modular, and computationally tractable operations rather than global analytical inversion. We formulate two hypotheses concerning structure-preserving constructibility and the transfer of well-posed local inverse realizations across faithful descriptions, together with an operational principle emphasizing causal, stable, robust, and predictably computable real-time realizations. Two propositions provide formal anchors: uniqueness and representation consistency of the local input–output inverse under an invertible linear term, and a constructibility result for a localized cubic element. Examples, including a Duffing oscillator with memory, illustrate both the possibilities and limitations of the framework. The resulting representation-aware perspective separates intrinsic invertibility from representation-dependent constructibility and computational cost.
Authors
- Augusto Sarti
Institutions
- Politecnico di Milano (IT)
Publication Details
- Journal
- Frontiers in Signal Processing
- Published
- 2026-11-16
- DOI
- https://doi.org/10.3389/frsip.2026.1951118
- Primary Topic
- Control Systems and Identification
- Type
- article
- Field-Weighted Citation Impact
- 0.00