Research Analyzer
← Back ICRA 2026

The Actuator Pre-Filtering Approach to Control-Coherent Koopman LQR for Robot Systems Interacting with Compliant Environment

Jasmine Terrones, Harry Asada

PDF

AI summary

Key figure (auto-extracted from paper)
An actuator pre-filtering technique enables globally linear Koopman modeling of non-autonomous robot systems, allowing real-time LQR control of complex, contact-switched dynamics.
Koopman operator Control-Coherent modeling Actuator pre-filtering Switched dynamics LQR control Robot-environment interaction

Problem

Robot systems interacting with environments experience switched dynamics that complicate real-time planning and control. Traditional Koopman operator theory cannot handle non-autonomous systems with exogenous control inputs, creating a gap for practical robotic applications.

Approach

The authors introduce an actuator pre-filter that replaces physical actuator dynamics with virtual linear filters, satisfying the mathematical conditions for Control-Coherent Koopman modeling. This transforms segmented nonlinear dynamics into a unified, globally linear state-space model suitable for linear control synthesis.

Key results

  • Actuator pre-filtering method formulated to satisfy Control-Coherent Koopman requirements
  • Globally linear Koopman model derived for a switched cart-pole system with compliant walls
  • Koopman LQR controller designed and validated for real-time wall-bouncing tasks
  • Simulation analysis quantifying the impact of pre-filter time constants on control performance

Why it matters

Provides a practical pathway for real-time optimal control of contact-rich robots by bypassing the limitations of traditional Koopman theory on non-autonomous systems.

Abstract

As a robot makes and breaks contact with environment surfaces, the equations of motion are switched. Task planning and real-time control become challenging as the system traverses multiple regions and switches the governing dynamics. This paper presents a modeling and real-time control methodology for such switched dynamical systems based on Koopman operator theory. Potentially, Koopman operators allow us to subsume segmented dynamics within a unified, globally linear model amenable for control analysis and synthesis. However, the original Koopman operators are not appliable to non-autonomous systems with exogenous input. A new method for converting robot dynamics to a Koopman- compatible model using actuator pre-filtering is presented and applied to the modeling and control of robots interacting with the environment. Specifically, an underactuated cart-pole robot bouncing against multiple walls is modeled as a Control- Coherent Koopman model and a Koopman LQR controller is designed for the wall-bouncing robot. Simulation experiments demonstrate the effectiveness of the method and investigates the effect of the actuator pre-filter parameter on control performance.

Index terms

Machine Learning for Robot Control Optimization and Optimal Control Underactuated Robots

Related papers