Coverage-Adaptive Random Walks for Fast Sensory Data Collection

Random walks in wireless sensor networks can serve as fully local, very simple strategies for sink motion that significantly reduce energy dissipation a lot but increase the latency of data collection. To achieve satisfactory energy-latency trade-offs the sink walks can be made adaptive, depending on network parameters such as density and/or history of past visits in each network region; but this increases the memory requirements. Towards better balances of memory/performance, we propose three new random walks: the Random Walk with Inertia, the Explore-and-Go Random Walk and the Curly Random Walk; we also introduce a new metric (Proximity Variation) that captures the different way each walk gets close to the network nodes over time. We implement the new walks and experimentally compare them to known ones. The simulation findings demonstrate that the new walks' performance (cover time) gets close to the one of the (much stronger) biased walk with memory, while in some other respects (partial cover time, proximity variation) they even outperform it. We note that the proposed walks have been fine-tuned in the light of experimental findings.