Distributed Koopman Learning with Incomplete Measurements

TL;DR

This paper introduces a distributed deep Koopman learning method enabling multiple agents with limited measurements to collaboratively identify nonlinear system dynamics, achieving accuracy comparable to centralized approaches.

Contribution

It develops a novel distributed algorithm combining Koopman theory, neural networks, and consensus, allowing agents to estimate global dynamics with partial data.

Findings

Achieves accurate system identification with incomplete measurements.

Performs comparably to centralized methods in simulations.

Demonstrates effectiveness on Lunar Lander environment.

Abstract

Koopman operator theory has emerged as a powerful tool for system identification, particularly for approximating nonlinear time-invariant systems (NTIS). This paper considers a network of agents with limited observation capabilities that collaboratively estimate the dynamics of an NTIS. A distributed deep Koopman learning algorithm is developed by integrating Koopman operator theory, deep neural networks, and consensus-based coordination. In the proposed framework, each agent approximates the system dynamics using its partial measurements and lifted states exchanged with its neighbors. This cooperative scheme enables accurate reconstruction of the global dynamics despite the absence of full-state information at individual agents. Simulation results on the Lunar Lander environment from OpenAI Gym demonstrate that the proposed method achieves performance comparable to the centralized deep Koopman learning with full-state access.