Date of Award
Doctor of Philosophy (PhD)
This dissertation considers two key problems in multi-agent systems: coordination (including both synchronization and desynchronization) and privacy preservation.
For coordination in multi-agent systems, we focus on synchronization/desynchronization of distributed pulse-coupled oscillator (PCO) networks and their applications in collective motion coordination. Pulse-coupled oscillators were originally proposed to model synchronization in biological systems such as flashing fireflies and firing neurons. In recent years, with proven scalability, simplicity, accuracy, and robustness, the PCO based synchronization strategy has become a powerful clock synchronization primitive for wireless sensor networks. Driven by these increased applications in biological networks and wireless sensor networks, synchronization of pulse-coupled oscillators has gained increased popularity. However, most existing results address the local synchronization of PCOs with initial phases constrained in a half cycle, and results on global synchronization from any initial condition are very sparse. In our work, we address global PCO synchronization from an arbitrary phase distribution under chain or directed tree graphs. More importantly, different from existing global synchronization studies on decentralized PCO networks, our work allows heterogeneous coupling functions and perturbations on PCOs' natural frequencies, and our results hold under any coupling strength between zero and one, which is crucial because a large coupling strength has been shown to be detrimental to the robustness of PCO synchronization to disturbances.
Compared with synchronization, desynchronization of PCOs is less explored. Desynchronization spreads the phase variables of all PCOs uniformly apart (with equal difference between neighboring phases). It has also been found in many biological phenomena, such as neuron spiking and fish signaling. Recently, phase desynchronization has been employed to achieve round-robin scheduling, which is crucial in applications as diverse as media access control of communication networks, realization of analog-to-digital converters, and scheduling of traffic flows in intersections. In our work, we systematically characterize pulse-coupled oscillators based decentralized phase desynchronization and propose an interaction function that is more general than existing results. Numerical simulations show that the proposed pulse based interaction function also has better robustness to pulse losses, time delays, and frequency errors than existing results.
Collective motion coordination is fundamental in systems as diverse as mobile sensor networks, swarm robotics, autonomous vehicles, and animal groups. Inspired by the close relationship between phase synchronization/desynchronization of PCOs and the heading dynamics of connected vehicles/robots, we propose a pulse-based integrated communication and control approach for collective motion coordination. Our approach only employs simple and identical pulses, which significantly reduces processing latency and communication delay compared with conventional packet based communications. Not only can heading control be achieved in the proposed approach to coordinate the headings (orientations) of motions in a network, but also spacing control for circular motion is achievable to design the spacing between neighboring nodes (e.g., vehicles or robots).
The second part of this dissertation is privacy preservation in multi-agent systems. More specifically, we focus on privacy-preserving average consensus as it is key for multi-agent systems, with applications ranging from time synchronization, information fusion, load balancing, to decentralized control. Existing average consensus algorithms require individual nodes (agents) to exchange explicit state values with their neighbors, which leads to the undesirable disclosure of sensitive information in the state. In our work, we propose a novel average consensus algorithm for time-varying directed graphs which can protect the privacy of participating nodes' initial states. Leveraging algorithm-level obfuscation, the algorithm does not need the assistance of any trusted third party or data aggregator. By leveraging the inherent robustness of consensus dynamics against random variations in interaction, our proposed algorithm can guarantee privacy of participating nodes without compromising the accuracy of consensus. The algorithm is distinctly different from differential-privacy based average consensus approaches which enable privacy through compromising accuracy in obtained consensus value. The approach is able to protect the privacy of participating nodes even in the presence of multiple honest-but-curious nodes which can collude with each other.
Gao, Huan, "Coordination and Privacy Preservation in Multi-Agent Systems" (2020). All Dissertations. 2718.