Certified Feedforward Tracking for Unknown Nonlinear Systems via Invertible Neural Networks

2026-08-05

AI Summary

This work proposes a data-driven feedforward control approach for partially observable, unknown nonlinear systems to achieve periodic trajectory tracking with theoretical performance guarantees. By employing an invertible neural network to construct a surrogate model of the system, the method circumvents the traditional non-convex inversion problem and its associated errors, recasting tracking error certification as a surrogate modeling task. Leveraging conformal prediction techniques, it provides, for the first time, a marginal probabilistic guarantee on feedforward tracking error under finite-sample conditions. The approach is experimentally validated on a DC motor load system with nonlinear friction, demonstrating high-accuracy and certifiable periodic trajectory tracking.

Abstract

In this paper, we address the certification of datadriven feedforward control for periodic tracking of unknown nonlinear systems under partial state measurements. To this end, we adopt an invertible neural network (INN) as a surrogate for the unknown system. This choice allows us to bypass solving a nonconvex inversion problem, eliminating the associated inversion errors and reducing tracking error certification to a surrogate modeling problem. We then apply conformal prediction to provide finite-sample probabilistic guarantees on the surrogate modeling error which, through the derived tracking error bound, yield marginal certificates on feedforward tracking error. Finally, we demonstrate the approach on a DC-motor-driven mechanical load with nonlinear friction.

Problem

• feedforward tracking
• unknown nonlinear systems
• certification
• partial state measurements
• periodic tracking

Innovation

• invertible neural networks
• feedforward control
• conformal prediction
• tracking error certification
• nonlinear systems