RDS-DeePC: Robust Data Selection for Data-Enabled Predictive Control via Sensitivity Score

TL;DR

This paper introduces RDS DeePC, a robust data selection method for Data Enabled Predictive Control that improves efficiency and outlier filtering using influence functions, applicable to nonlinear systems, with demonstrated high accuracy and performance.

Contribution

The paper presents a novel influence function-based data selection framework for DeePC, enabling automatic outlier filtering and computational efficiency without data quality labels, extended to nonlinear systems.

Findings

Achieves 94-97% clean data selection.

Maintains comparable or better tracking with 20% data corruption.

Effective for both linear and nonlinear systems.

Abstract

Data Enabled Predictive Control (DeePC) is an established model free approach to predictive control, but it faces two open challenges: computational complexity that scales cubically with dataset size and performance degradation when data are corrupted. This paper introduces Robust Data Selection DeePC (RDS DeePC), a framework that addresses both obstacles through influence function analysis. We derive a sensitivity score quantifying the leverage each trajectory segment exerts on the optimization solution and prove that high sensitivity segments correspond to outliers while low sensitivity segments represent consistent data. Selecting low sensitivity segments thus yields both computational efficiency and automatic outlier filtering without requiring data quality labels. For nonlinear systems, we extend the framework via a two stage online selection approach accelerated by the LiSSA algorithm. Experiments on four systems of increasing complexity including a DC motor, an inverted pendulum, a planar quadrotor UAV tracking a figure 8 trajectory, and a kinematic bicycle vehicle following a figure 8 path demonstrate that RDS DeePC achieves 94 to 97 percent clean data selection and comparable or better tracking performance under 20 percent data corruption.