Consensus-Driven Particle Swarm Optimization for Cooperative Resource Allocation in Heterogeneous Distributed Sensor and Actuator Networks

Abstract

Distributed sensor and actuator networks operate under constraints of energy, bandwidth, latency, and partial observability, while often facing heterogeneous node capabilities and time-varying connectivity. Cooperative resource allocation in such networks requires balancing local operational goals with system-level coordination without imposing heavy centralized infrastructure. Particle Swarm Optimization has been applied to nonconvex resource assignment because it is derivative-free and parallelizable, yet classical forms are communication-blind and sensitive to network fragmentation. This paper examines a consensus-driven variant of Particle Swarm Optimization designed for cooperative allocation across heterogeneous nodes that exchange low-dimensional states with neighbors. The method integrates a mixing process that steers particles toward neighborhood barycenters, while preserving exploratory diversity necessary to escape poor local allocations. The study formulates a network-regularized objective that captures coupling through shared actuators and congested links, and examines stability under general conditions on mixing weights, inertia, and noise. A Lyapunov-style argument outlines how consensus error and swarm potential are jointly contracted in expectation under mild spectral-gap assumptions on the communication graph. The computational and communication footprints are characterized, with attention to asynchronous message arrivals, quantized signaling, and privacy of local cost information. Numerical protocols are described to probe sensitivity to network sparsity, heterogeneity, and measurement noise. The discussion offers guidance on parameterization for typical embedded deployments and interprets empirical behavior in terms of dynamic regret and allocation smoothness. The presentation aims to be technically complete while remaining conservative in claims about performance under unmodeled disturbances.