Cooperative output regulation for networks of hyperbolic systems with unknown leader and disturbances

Read full text Summarise with VIRAT
Authors: Tarik Enderes
Cited by: 0
References: 127

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The aim of this work is to solve the cooperative output synchronization problem for networks of systems with heterodirectional hyperbolic dynamics in the presence of unknown leaders and disturbances. Therein, the agents are all governed by a partial differential equation (PDE) with a temporal and spatial component, making them infinite-dimensional. The agent outputs are to track a reference signal that is generated by a leader with unknown dynamics asymptotically while simultaneously rejecting their individual unknown disturbances. These control outputs may depend on the agent states at the boundaries as well as point-wise and distributed within the spatial domain. This is achieved cooperatively, wherein the agent communication is restricted to a network described by a directed graph, through boundary actuation of the PDE dynamics. Both the leader and disturbances are described by finite-dimensional ordinary differential equations (ODEs) called the signal models, which allow to describe constant, polynomial and harmonic signals for a non-trivial output regulation problem. The parameters of the signal models generating these external signals are unknown to the followers, but are available to the leader. To address the issue of unknown leader and disturbance dynamics, cooperative signal model observers are used that allow the agents to reconstruct the signal model parameters online from a networked communication by solving a synchronization problem. The generated estimates are then used to design an adaptive controller that is appropriate for the respective setup, leading to a fully distributed solution. Using dynamic regulators with time-dependent parameters and adapting the control gains at run-time results in time-varying closed-loop dynamics, which impedes the stability analysis. Further, this approach raises the need to solve some boundary value problems (BVPs) online. Solvability issues of certain equations may arise during the transient process, potentially leading to undefined behavior. Solutions to these problems are proposed in this thesis w.r.t. the considered controller design method. The adaptive approach using the cooperative signal model observer is introduced by first considering a network of uncertain agents with ODE dynamics. For this purpose, a state feedback controller based on the internal model principle is presented. Therein, a cooperative internal model that is driven by the diffusively coupled tracking error is continuously updated using the estimates of the signal model dynamic matrix. By stabilizing the extended system consisting of the agent dynamics and the adaptive cooperative internal model, a regulator is obtained that ensures robust cooperative output regulation. To resolve solvability issues, a sample-and-hold strategy for the adaptation of the control gains is proposed, which is also used in the subsequent setups. This method permits differences between the agents in terms of model uncertainties, but not in regard to the nominal system parameters. The results are validated by a numerical simulation of four uncertain agents as well as a practical experiment using mobile robots. After introducing the adaptation strategy in the ODE case, the adaptive cooperative internal model is then applied to the robust cooperative output regulation problem for a network of hyperbolic PDE agents. For this, a regulator is designed using the backstepping method and a decoupling of the ODE and PDE states. The latter requires to solve a BVP with parameters that are time-dependent due to the adaptive strategy and therefore need to be solved at run-time. An efficient approximate solution to achieve this is proposed. The stability of the time-varying closed-loop PDE-ODE dynamics is proven rigorously using the method of characteristics. It is verified that this solves the robust cooperative output regulation problem. This method is demonstrated using an academic example of four uncertain agents. Since the internal model principle is only applicable if the agents are homogeneous in the nominal case, another method is presented to address a fully heterogeneous multi-agent system (MAS). This problem is solved by a local adaptive feedforward controller, which is driven by adaptive cooperative reference and disturbance observers. Thereby, the adaptive cooperative disturbance observer estimates the disturbance state and dynamics as well as the agent state. This amounts to hyperbolic PDE-ODE equations with time-varying system parameters. Thus, stabilizing the observer error dynamics with the backstepping approach raises similar challenges as the previous setup but requires distinctly different solutions. By combining the observers with an adaptive feedforward controller, which requires an online approximation of the solution of the regulator equations, this regulator ensures cooperative output regulation. An academic example using four distinct agents illustrates these results as well.

Fields of Study & Subjects

Stability and Controllability of Differential Equations Distributed Control Multi-Agent Systems Neural Networks Stability and Synchronization

Publication details

Publication date
1 January 2026
Language
EN
Document type
Thesis