Distributed H-infinity Tracking Control for Discrete-Time Multi-Agent Systems with a High-Dimensional Leader

Guanghui Wen, Valery A. UgrinovskiiarXiv 2013 · OpenReview

TL;DR

This paper tackles the challenge of making a group of agents (followers) accurately track the movements of a more complex leader agent, even when the leader's full state isn't directly visible to all followers and the followers themselves have noisy dynamics. They propose a novel distributed control strategy that combines local controller feedback with a distributed state estimator, achieving robust tracking performance.

What Problem Does This Paper Address?

Coordinating multiple agents to follow a leader is a fundamental problem in robotics and control systems. However, existing methods often assume that all agents have identical dynamics and that the leader's state is fully observable. This paper addresses a more realistic scenario:

  • High-Dimensional Leader: The leader agent has a more complex internal state than the followers.
  • Partial Observability: Only a subset of followers can sense some information about the leader's state.
  • Noisy Followers: The follower agents' dynamics are subject to external disturbances.
  • Distributed Control: Each follower must make control decisions based only on information from its immediate neighbors and its own state.

The goal is to design a control system that ensures all followers can accurately track the leader's trajectory while maintaining a guaranteed level of performance (H-infinity norm) in the presence of these challenges.

Methods

The authors propose a distributed dynamic tracking protocol for each follower agent. This protocol has two key components:

  1. Neighbor-Based Local Controller: This part of the controller uses information from the follower's own state and the states of its neighboring followers to generate a control signal.
  2. Distributed State Estimator: This component is crucial for dealing with the partial observability of the leader. It estimates the leader's state using information from neighbors and the limited output information available from the leader.

The core of their approach involves designing these components such that the closed-loop system of followers achieves a desired H-infinity tracking performance. This is achieved by converting the distributed problem into a set of decoupled H-infinity control problems for individual followers, whose solvability is then analyzed using linear matrix inequalities (LMIs). The communication topology is assumed to have a directed spanning tree from the leader to the followers, and the follower-to-follower communication graph is undirected.

Datasets

Not discussed in the paper. The paper focuses on theoretical analysis and simulation results based on a defined system model.

Key Insights and Ideas

  • Bridging the Gap in Complexity: The paper successfully extends leader-following control to scenarios where the leader's dynamics are significantly more complex than the followers', a gap not well-addressed by prior work on identical agent systems.
  • Distributed Estimation for Robustness: The integration of a distributed state estimator is key to handling the partial observability of the leader, allowing followers to infer the leader's state without direct access.
  • H-infinity Guarantees: The use of H-infinity control theory provides a rigorous framework for guaranteeing a specific level of performance and robustness against disturbances, ensuring that the tracking error remains bounded.
  • Topology Matters: The paper highlights how the structure of the communication network (specifically, the directed spanning tree from the leader and the undirected follower graph) influences the solvability and performance of the tracking problem.

Key Results

The paper presents theoretical conditions for achieving distributed H-infinity leader-following tracking.

  • Theorem 1: Establishes necessary and sufficient conditions for solvability by converting the distributed problem into a set of N-1 independent H-infinity control problems, each characterized by the eigenvalues of the communication topology matrix.
  • Theorem 2: Provides sufficient conditions for designing the feedback gain matrix (F) using LMIs, enabling the construction of the controller and estimator.
  • Simulations: Numerical simulations demonstrate the effectiveness of the proposed protocol. Figures 2-4 show that the followers' states and the estimated leader states converge to the leader's states. Figure 5 illustrates the consensus tracking error converging to zero in the absence of disturbances. Figure 6 shows that the H-infinity performance level is achieved under disturbance conditions.

Limitations and Future Work

  • Limitations:
    • The paper assumes a fixed communication topology.
  • The leader's dynamics are assumed to be known to the followers, though its initial condition is unknown.
  • The follower-to-follower communication graph is assumed to be undirected.
  • Future Work: The authors suggest extending their work to:
  • Multi-agent systems with time-varying topologies.
  • Systems with high-dimensional leaders and time-varying topologies.
  • Multi-agent systems with nonlinear dynamics.

Summary updated

Abstract

This paper considers the distributed H-infinity leader-following tracking problem for a class of discrete time multi-agent systems with a high-dimensional dynamic leader. It is assumed that output information about the leader is only available to designated followers, and the dynamics of the followers are subject to perturbations. To achieve distributed H-infinity leader-following tracking, a new class of control protocols is proposed which is based on the feedback from the nearest neighbors as well as a distributed state estimator. Under the assumptions that dynamics of the leader are detectable and the communication topology contains a directed spanning tree, sufficient conditions are obtained that enable all followers to track the leader while achieving a desired H-infinity leader-following tracking performance. Numerical simulations illustrate the effectiveness of the theoretical analysis.
BibTeX
@misc{wen2013distributed,
  author = {Guanghui Wen and Valery A. Ugrinovskii},
  title = {{Distributed H-infinity Tracking Control for Discrete-Time Multi-Agent Systems with a High-Dimensional Leader}},
  year = {2013},
  eprint = {1308.6709},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1308.6709},
}
Figure