Abstract
This paper proposes a novel distributed event-triggered algorithmic solution to the multi-agent average consensus problem for networks whose communication topology is described by weight-balanced, strongly connected digraphs. The proposed event-triggered communication and control strategy does not rely on individual agents having continuous or periodic access to information about the state of their neighbors. In addition, it does not require the agents to have a priori knowledge of any global parameter. We show that, under the proposed law, events cannot be triggered an infinite number of times in any finite period (i.e., no Zeno behavior), and that the resulting network executions provably converge to the average of the initial agents' states exponentially fast. We also provide weaker conditions on connectivity under which convergence is guaranteed when the communication topology is switching. Simulations illustrate our results.
Introduction
In multi-agent systems, average consensus is a fundamental task where every agent in a network agrees on the average of their initial values. This process is essential for distributed sensing, load balancing, and formation control. Traditional approaches typically require agents to continuously or periodically poll their neighbors. However, in real-world networked systems, constant communication is costly, draining batteries and consuming bandwidth.
While event-triggered control—where agents only communicate when necessary—offers a solution, existing methods often require global network knowledge or assume undirected communication. This paper introduces a fully distributed solution for weight-balanced digraphs that requires no global knowledge and guarantees the avoidance of "Zeno behavior" (infinite communication in finite time).
Methodology
The authors employ a Lyapunov-based design to derive triggering laws. They define a "disagreement function" to measure the network's distance from consensus and determine the maximum error an agent can tolerate before a broadcast is required.
Triggering Logic
Each agent monitors the difference between its current state and its last broadcast state. When this error exceeds a threshold derived from local neighbor data, an event is triggered. To prevent Zeno behavior, the authors incorporate a "waiting period" ($\epsilon_i$) and specific triggers to handle discontinuities when receiving new information.
The algorithm ensures that agents only broadcast their state $x_i(t)$ when a local condition $f_i(e_i(t)) > 0$ is met, or when specific information updates from neighbors necessitate a re-broadcast to maintain system stability.
Key Results
The proposed method was evaluated through numerical simulations of a 5-agent network, comparing it against standard periodic Laplacian consensus and previous state-of-the-art methods.
Performance and Efficiency
Mathematical proofs and simulations confirm that the network converges to the average of initial states at an exponential rate. Crucially, the proposed method is significantly more communication-efficient.
As shown in the figure above, the proposed algorithm (solid blue line) achieves consensus with a fraction of the broadcasts required by previous methods (dashed grey line), demonstrating superior efficiency in bandwidth usage.
Robustness
The paper also establishes conditions for consensus under switching topologies, where network links may break and reform, provided the network remains "jointly" connected over time.
Conclusion and Future Work
This work provides a robust, distributed framework for average consensus on directed graphs without requiring global parameters. While effective for single-integrator dynamics, future research could extend these results to more complex agent physics (e.g., double-integrator dynamics) and account for non-ideal communication factors like delays or packet loss. Additionally, integrating distributed weight-balancing schemes could broaden the applicability to general strongly connected digraphs.
Summary updated