Towards Distributed Optimal Movement Strategy for Data Gathering in Wireless Sensor Networks

Abstract

We demonstrate Towards Distributed Optimal Movement Strategy for Data Gathering in Wireless Sensor Networks that under various scenarios our distributed optimal movement strategy can lead to about two times lower loss rates than a standard random walk strategy. In particular, our strategy results in cost savings of up to 70 per cent for the deployment of multiple collectors to achieve the target data loss rate compared to the standard random walk strategy.

For the data collection process, a sink (or base station) generates periodic query packets or collector packets to collect certain information of interest from sensor nodes or each sensor, instead informing the sink node directly about its observed data or events.

In addition, mobile sinks (agents), e.g. data mules, can move around the sensor field and collect information from each sensor. These metrics are suitable for delivery of one-shot information or search / quest. In contrast, in this paper, we look at the problem of collecting data from a different but important perspective using the random walk agent(s).

It is therefore more important to measure how many data can be collected before being lost due to limited buffer space, when the sink generates periodically query packets or collector packets moving across the network in a random walk fashion to collect measured data or its aggregated / compressed version from sensor nodes.

In this regard, many research works are based on traveling salesman problem, i.e. finding a Hamiltonian cycle (a NP-hard problem), or its variants, and their solutions are generally solvable only globally or require global network information such as sensor location / distance information. In this paper, without such global information, we focus on a class of Markovian random walks.

Conclusion

We have developed an analytical framework to evaluate the probability of network loss as a performance metric for different mobile collector Markovian movement strategies moving over a graph (or network) for data harvesting in WSNs.