Energy-efficient Real-time Communication in Multi-hop Low-power Wireless Networks

Abstract

Low-power wireless holds the promise of improving reliability and reducing costs in control applications. The key challenge in achieving these goals is to deliver packets within real-time deadlines across devices with limited energy. Existing approaches either can not provide end-to-end guarantees due to a fully localized operation, or hardly scale as they are sensitive to dynamic changes in the network state. Our key insight is that a fully global approach can overcome these limitations by being agnostic to the current network state. To substantiate this claim, we build Blink, a real-time low-power wireless protocol that provides hard guarantees on end-to-end packet deadlines, scales to large multi-hop networks, and seamlessly handles dynamic changes in network state and real-time requirements. We achieve this by leveraging an existing best-effort protocol that uses only flooding for communication, and by designing novel scheduling algorithms based on the earliest deadline first (EDF) policy. Using a dedicated priority queue data structure, we demonstrate a viable implementation of our algorithms on resource-constrained devices. Results from a 94-node testbed and an instruction-level emulator show that Blink: (i) meets almost 100% of packet deadlines, missing only a few due to packet losses over the wireless channel; (ii) keeps the network-wide energy consumption to a minimum; and (iii) schedules 200 real-time packet streams in less than 80 milliseconds on a 16-bit 8 MHz microcontroller.