Fast average gossiping under asymmetric links in WSNS

Wireless Sensor Networks are a recent technology where the nodes cooperate to obtain, in a totally distributed way, certain function of the collected data. An important example of these distributed processes is the average gossip algorithm, which allows the nodes to obtain the global average by only using local data exchanges. This process is traditionally slow, but can be accelerated by introducing geographic information or by exploiting the broadcast nature of the wireless medium. However, when a gossip protocol utilizes long geographic routes or broadcast communications, its convergence is not easily guaranteed due to asymmetry in communications. Alternatively, we propose an asymmetric version of the gossip algorithm that exploits residual information involved in each asymmetric exchange. Our asymmetric gossip algorithm achieves convergence faster than existing studies in the related literature. Numerical results are presented to show clearly the validity and efficiency of our approach.