Controlled Synchronization of Coupled Pendulums by Koopman Model Predictive Control

TL;DR

This paper introduces a real-time Koopman Model Predictive Control method to synchronize a chain of coupled nonlinear pendulums actuated from one end, demonstrating control of various equilibrium and periodic states.

Contribution

It presents a novel control approach combining Koopman operator theory with MPC for synchronized control of nonlinear pendulum chains with a single actuator.

Findings

Successfully synchronized pendulums to stable and unstable equilibria.

Achieved synchronization to periodic orbits.

Real-time implementation demonstrated.

Abstract

We propose and experimentally demonstrate a feedback control method that allows synchronizing the motion of a chain of several coupled nonlinear oscillators actuated through one end of the chain. The chain considered in this work is a one-dimensional array of pendulums pivoting around a single axis and interacting with adjacent pendulums through torsion springs; the array is actuated using a single torque motor attached to one of the two boundary pendulums. This represents a mechanical realization of the Frenkel-Kontorova model { a spatially discrete version of a sine-Gordon equation describing (nonlinear) waves. The main challenges of controlling these systems are: high order (the number of pendulums can be high), nonlinear dynamics, and (as we set the problem here) only one actuator. The presented problem of synchronization of motion is a special case of the problem of reference tracking, where all pendulums reach a common point or a trajectory. In particular, we demonstrate synchronization to a stable equilibrium (all pendulums downward), unstable equilibrium (all pendulums upward), and a periodic orbit (all pendulums revolving). We use the Koopman Model Predictive Control (KMPC) that constructs a linear predictor of the nonlinear system in a higher-dimensional lifted space and uses the predictor within a classical linear MPC, thereby maintaining low computational cost that allows for a real-time implementation, while taking into account the complex nonlinear dynamics.