A novel dynamic event-triggered mechanism for dynamic average consensus

Tao Xu, Zhisheng Duan, Guanghui Wen, Zhiyong Sun

TL;DR

This paper introduces a novel distributed algorithm for dynamic average consensus (DAC) that uses an event-triggered communication strategy. Unlike previous methods, it guarantees a positive minimum time between communication events without sacrificing estimation accuracy, making it more practical for real-world systems with limited bandwidth.

What Problem Does This Paper Address?

The dynamic average consensus (DAC) problem involves a group of agents trying to estimate the average of time-varying reference signals, each known only to a subset of agents. This is crucial for applications like distributed formation control and state estimation. While many algorithms exist, they often suffer from two main limitations:

  1. Initialization Sensitivity: Many algorithms require specific initial conditions (e.g., the sum of initial states being zero), which is impractical if communication links break and agents form subgroups that need re-initialization.
  2. Continuous Communication: Traditional methods demand constant data exchange between agents, straining limited wireless bandwidth and computational resources.

To address these, event-triggered mechanisms were proposed, where communication only happens when a specific "event" occurs. However, existing event-triggered DAC solutions still face challenges: some are initialization-sensitive, others require global network information, and crucially, many either don't guarantee a positive and uniform minimum interval between events (leading to potential Zeno behavior or impractical rapid sampling) or only bound the estimation error instead of driving it to zero. This paper aims to bridge this gap by developing a robust, distributed, event-triggered DAC algorithm that is initialization-free, requires no global information, and ensures both zero estimation error and a guaranteed positive minimum inter-event interval.

Methods

The paper proposes a three-pronged approach:

  1. Robust Adaptive Distributed DAC Algorithm: This algorithm estimates the average reference signal. A key feature is its robustness to initialization, meaning it works even if initial states are arbitrary. It uses a distributed adaptive input ui(t) that relies on local information from neighbors.
  2. Novel Dynamic Event-Triggered Mechanism: This is the core innovation. Instead of a simple threshold, it uses a dynamic triggering function fi(t) that is itself updated over time. This mechanism determines when an agent should broadcast its state to neighbors. The novelty lies in its ability to guarantee a positive and uniform minimum inter-event interval for all agents, preventing the Zeno phenomenon (infinitely many events in finite time) and ensuring practical sampling rates.
  3. Composite Adaptive Update Law: To manage the adaptive gains (cij(t)) used in both the DAC algorithm and the triggering mechanism, a composite adaptive law is introduced. This law, along with a σ-modification technique, helps ensure zero estimation error and allows the algorithm to operate without any global network information.

Datasets

Not discussed in the paper. The evaluation is based on numerical simulations with synthetically generated reference signals.

Key Insights and Ideas

  • Guaranteed Minimum Inter-Event Interval: The most significant insight is the design of a dynamic triggering function that mathematically guarantees a positive and uniform minimum time between communication events. This is a critical step towards practical event-triggered systems, moving beyond just excluding Zeno behavior.
  • Zero Estimation Error with Event-Triggering: The paper demonstrates that it's possible to achieve perfect estimation (zero error) even with intermittent communication, a challenging feat in distributed consensus problems.
  • Initialization-Free Operation: By making the algorithm robust to initial conditions, it becomes more applicable to real-world scenarios where network disruptions can occur, requiring agents to adapt without needing a global reset.
  • No Global Information Required: The reliance solely on local information from neighbors simplifies implementation and enhances scalability, as agents don't need to know the entire network topology or the total number of agents.
  • Trade-off in Triggering: The paper acknowledges a trade-off between the speed of estimation convergence and the triggering performance (i.e., the size of the minimum inter-event interval). A more aggressive triggering can lead to faster convergence but potentially smaller inter-event intervals, and vice-versa.

Key Results

  • Theorem 1: Proves that the proposed algorithm guarantees that the estimation error xi(t) converges to zero as time goes to infinity. It also shows that the adaptive gains cij(t) and their compensation terms ĉij(t) converge to the same positive constant.
  • Theorem 2: Proves that the novel dynamic event-triggered mechanism ensures the existence of a positive and uniform lower bound for the inter-event intervals for all agents at all times.
  • Simulations: Numerical simulations with 8 agents divided into two subgroups, and with simulated communication link failures, demonstrate the effectiveness of the algorithm. The results show accurate estimation, convergence of errors to zero, and the dynamic triggering functions behaving as expected, with statistics confirming positive minimum inter-event intervals.

Limitations and Future Work

  • Disturbance Handling: The authors mention that the effect of disturbances on the distributed DAC solution design and analysis will be studied in their future work.
  • Complexity of Adaptive Gains: While the adaptive gains converge, their dynamic update law might introduce computational overhead. The paper focuses on theoretical guarantees, and practical implementation efficiency could be a further research area.
  • Generalization to Higher Dimensions: While the paper states results are extendable to arbitrary dimensions for reference signals, the simulations are presented for scalar signals. Further exploration of multi-dimensional cases might be beneficial.

Summary updated

Abstract

This paper studies a challenging issue introduced in a recent survey, namely designing a distributed event-based scheme to solve the dynamic average consensus (DAC) problem. First, a robust adaptive distributed event-based DAC algorithm is designed without imposing specific initialization criteria to perform estimation task under intermittent communication. Second, a novel adaptive distributed dynamic event-triggered mechanism is proposed to determine the triggering time when neighboring agents broadcast information to each other. Compared to the existing event-triggered mechanisms, the novelty of the proposed dynamic event-triggered mechanism lies in that it guarantees the existence of a positive and uniform minimum inter-event interval without sacrificing any accuracy of the estimation, which is much more practical than only ensuring the exclusion of the Zeno behavior or the boundedness of the estimation error. Third, a composite adaptive law is developed to update the adaptive gain employed in the distributed event-based DAC algorithm and dynamic event-triggered mechanism. Using the composite adaptive update law, the distributed event-based solution proposed in our work is implemented without requiring any global information. Finally, numerical simulations are provided to illustrate the effectiveness of the theoretical results.
BibTeX
@article{xu2024novel,
  author = {Tao Xu and Zhisheng Duan and Guanghui Wen and Zhiyong Sun},
  title = {{A novel dynamic event-triggered mechanism for dynamic average consensus}},
  journal = {Automatica},
  year = {2024},
  url = {https://openreview.net/forum?id=cVMHjYL7cM},
}
Figure