Self-learning and self-adaptive framework for supporting high reliability and low energy expenditure in WSNs

In this paper we proposed, designed and evaluated a novel decentralized and self-learning framework to support both high reliability and energy-efficiency for periodic traffic applications in WSNs. Our autonomous framework comprises three main components: estimation and identification of periodic flows, dynamic wakeup-sleep scheduling and asynchronous channel hopping. With asynchronous channel hopping the frequency hopping pattern is determined by each source node autonomously, and forwarders have to identify and follow the pattern. We also propose a light and efficient controller to eliminate the collision caused by multi-flow overlap at forwarders. We present design and evaluation of our autonomous framework using realistic trace-based simulation. The results show that our asynchronous channel hopping solution improves the packet reception rate compared to the single channel solutions without the need of an expensive signaling and time synchronization overhead. We also show that with this scheme the average energy consumption yields a $$\approx $$≈ 50 % lower than the single channel solutions. Furthermore, we analyze in detail the energy consumption characteristics of our autonomous framework when operated with a popular transceiver, the ChipCon CC2420 and Texas Instruments MSP430 Microcontroller. We analyze how much various factors contribute to the overall energy consumption. These insights provide valuable guidance on where to start with any effort geared towards saving energy.