Consensus of second-order multi-agent systems with time-varying delays and antagonistic interactions

Bo Hou, Fuchun Sun, Hongbo Li, Guangbin LiuTsinghua 2015 · DOI

Abstract: This study investigates the consensus problem of second-order multi-agent systems subject to time-varying interval-like delays. The notion of consensus is extended to networks containing antagonistic interactions modeled by negative weights on the communication graph. A unified framework is established to address both the stationary and dynamic consensus issues in sampled-data settings. Using the reciprocally convex approach, a sufficient condition for consensus is derived in terms of matrix inequalities. Numerical examples are provided to illustrate the effectiveness of the proposed result.

Introduction

In the field of Multi-Agent Systems (MAS), "consensus" typically refers to a state where a group of autonomous agents—such as robots, drones, or sensors—reaches an agreement on a specific variable, like position or velocity. While much of the existing literature assumes a cooperative environment where agents share information through positive weights, real-world scenarios often involve antagonistic interactions. These are common in social networks, competitive markets, or adversarial robotics, where one agent's influence on another is negative.

Furthermore, communication in digital networks is rarely instantaneous. Time-varying delays are ubiquitous and can significantly destabilize a system. This paper addresses the challenge of achieving consensus in second-order systems (where control affects acceleration) under the simultaneous presence of antagonistic behavior—modeled via signed graphs—and interval-bounded, time-varying delays.

Methodology

To address these complexities, the researchers established a unified mathematical framework using several advanced control theory tools:

  • Signed Graphs: Standard graphs are extended to include negative edge weights, representing antagonistic relationships between agents.
  • Sampled-Data Setting: The system operates in discrete time steps, reflecting the behavior of digital controllers.
  • Unified Error Framework: The authors transformed two distinct consensus types—stationary (convergence to a fixed point) and dynamic (convergence to a shared moving trajectory)—into a single stabilization problem. By defining error as the difference between an agent and a reference, the goal becomes stabilizing a linear system to zero.
  • Lyapunov-Krasovskii Functional (LKF) & Reciprocally Convex Approach: To prove stability despite delays, a complex energy-like function (LKF) is used. The reciprocally convex approach is applied to reduce mathematical "conservativeness," providing a tighter and more accurate bound for the maximum allowable delay.
Figure 3: Communication topology of the four-agent system, where dashed lines represent antagonistic interactions.

One critical insight from this framework is that for consensus to be achievable in a signed graph, there must be at least one cooperative link between an agent and its neighbors. An agent that only interacts antagonistically cannot reach an agreement with the group.

Key Results and Validation

The proposed methodology provides a set of Linear Matrix Inequalities (LMIs) that serve as a "recipe" for guaranteed consensus. The effectiveness of these conditions was validated through numerical simulations of a four-agent system.

  • Stationary Consensus: Using a damping gain of $\kappa_s = 2$, the system successfully reached consensus even with random time-varying delays of up to 153 sampling periods.
  • Dynamic Consensus: Achieving consensus while moving proved more challenging. With a control gain of $\kappa_d = 2$, the maximum allowable delay was found to be 39 sampling periods.
Figure 5: Dynamic consensus results showing (a) position and (b) velocity convergence for four agents despite antagonistic links and delays.

As shown in the simulation plots, despite initial fluctuations caused by the antagonistic "push-pull" and communication delays, the agents' positions and velocities eventually converge to a common trajectory, demonstrating the robustness of the proposed control law.

Conclusion

This research successfully extends consensus theory to more hostile and realistic environments. By treating consensus as a stabilization problem and utilizing the reciprocally convex approach, the authors provide a practical method for designing controllers for second-order MAS with delays and negative interactions.

Future work may address current limitations, such as the assumption of identical delays across all agents (uniform delays) and the use of a fixed communication topology. Exploring non-uniform delays and switching topologies—where agents change their relationships or move in and out of range—remains an important frontier for large-scale multi-agent coordination.

Summary updated

Abstract

This study investigates the consensus problem of second-order multi-agent systems subject to timevarying interval-like delays. The notion of consensus is extended to networks containing antagonistic interactions modeled by negative weights on the communication graph. A unified framework is established to address both the stationary and dynamic consensus issues in sampled-data settings. Using the reciprocally convex approach, a sufficient condition for consensus is derived in terms of matrix inequalities. Numerical examples are provided to illustrate the effectiveness of the proposed result.
BibTeX
@inproceedings{hou2015consensus,
  author = {Bo Hou and Fuchun Sun and Hongbo Li and Guangbin Liu},
  title = {{Consensus of second-order multi-agent systems with time-varying delays and antagonistic interactions}},
  booktitle = {Tsinghua},
  year = {2015},
  doi = {10.1109/tst.2015.7085634},
}
Figure