Swarm Intelligence for Resilient Network Coverage: a Coagulation-Inspired Paradigm

Wireless sensor networks are becoming widely used to sense large areas, with network nodes providing coverage to track and monitor phenomena of interest. The growing demand for these sensing systems has highlighted the need for solutions that ensure continuous area coverage, even when external factors disrupt service. In particular, node failures can leave parts of the area uncovered, creating holes in coverage. Existing solutions for detecting and repairing, or healing, such holes often rely on unrealistic assumptions, like the need for node mobility, redundancy, and that nodes will not fail during the repair process. To address these limitations, new solutions focus on using external nodes to heal coverage holes, thus eliminating these assumptions. For these solutions towork, external nodes must cooperate autonomously and offer scalability, flexibility to adapt to different networks, and robustness to handle failures. These traits are naturally achieved by applying the swarm intelligence paradigm to nodes’ control logic. In this thesis, we tackle the problem of hole detection and healing by proposing three swarm intelligence-based algorithms that use resource-limited agents to temporarily restore sensing in the network. We argue that controlling these external nodes, or swarm agents, with logic inspired by the blood coagulation process offers significant advantages. In the first algorithm, agents are modeled as artificial platelets, following the adapted biological rules of activation, adhesion, and cohesion. Once activated, the swarm agents navigate the network using cues from nearby nodes to locate the nearest hole and adhere to its borders at locally optimal positions. These agents then attract the rest of the swarm, ensuring cohesion until coverage restoration is complete. We also propose the Discrete variant of this algorithm, which adapts to scenarios with extremely quantized and temporally sparse agents’ perception, addressing limited swarm functionality in complex environments. Building on insights from both approaches, we introduce the BHDH algorithm, designed to accelerate the healing process while minimizing resource use, incorporating a battery model to bridge the gap with real-world scenarios and extend the solution’s lifetime. We validate the proposed algorithms using HDHSim, a discrete-time simulator we developed for hole detection and healing problems. The approaches demonstrate efficient coverage hole detection and service restoration, regardless of hole size, shape, location, or agents’ sensing capabilities, while showing exceptional robustness to agent failure. Despite the limited perception, the Discrete variant achieves similar performance to the first algorithm while providing faster healing, though requiring more agents. The BHDH algorithm combines the strengths of both solutions to deliver the fastest healing with minimal agents, aided by improved resource management through energy information. Our comparisons with state-of-the-art solutions show that our algorithms, especially BHDH, outperform them, highlighting the effectiveness of physiological phenomena as inspiration for swarm control.