Echo state neural network-assisted mobility-aware seamless handoff in mobile WSNs

Abstract

Strong mobility in Wireless Sensor Networks (WSNs) may disrupt an existing link between two communicating nodes. Because many Medium Access Control (MAC) protocols cannot adapt to the strong mobility of nodes, facing a decaying link, a node usually chooses to continue its data communication until the current link is terminated, and then searches for a new relay node to build a better connection. However, this will cause serious transmission latency. To tackle this problem, this paper proposes an echo state neural network-assisted mobility-aware seamless handoff method, which enables a node to handoff communication seamlessly to a more reliable link while maintaining communication over the current connection, and the current connection will be interrupted only after a new link is set up. To demonstrate the superiority of handoff, by taking the cattle monitoring as an example where nodes move at a constant speed, this paper designs a static-oriented Beacon-Initiated MAC (BI-MAC) protocol, develops a distance threshold as the handoff initiation signal, predicts future locations of nodes with a machine learning-based Echo State Network (ESN), formulates a mobility-aware seamless handoff module on top of the BI-MAC, and compares the data communication latency using and not using the handoff. Both the analytical and NS2 simulation results show that the time to set up a new link is much longer than the handoff latency. The one-hop latency can be reduced by at least 42% with the handoff support. This figure can further grow with the increase in duty cycle, network density, and size of multi-hop links.