Collision forecasting: A low-power MAC with traffic and power shaping

摘要

In an important class of mobile networked applications, it is not computation or sensing, but radio communications that dominate the wireless nodes' power budget. While it is often required that a predefined minimum network lifetime be guaranteed at deployment time, it is nontrivial to control the power consumption of the radio communication subsystem when the network size may change over time.We present the Asynchronous Random Schedules with Collision Forecast (ARS/CF) protocol, a low-power MAC for infrastructureless single-hop networks, with MAC-level support for prescribing a power budget. Instead of relying on channel sampling or common communication schedules, ARS/CF uses randomized local transmission schedules, which are shared with neighbors within the collision domain. Collisions are forecasted by examining the transmission times in the local as well as in the neighbors' schedules. When a collision is imminent, the corresponding transmissions and receiver wake-ups are canceled ahead of time, thereby avoiding unnecessary waste of energy on both the sender and receiver side. ARS/CF controls energy consumption by dynamically adjusting transmission rates. Given a pre-set power budget, as well as an estimate of the typical number of nodes in the collision domain, we derive a closed form solution for the optimal protocol parameters, and provide guarantees that the average power draw will stay within the predefined envelope even if the actual neighborhood size is different from the estimate. We provide theoretical as well as experimental results on the performance of ARS/CF and show that it outperforms CSMA-based asynchronous low-power MACs under periodic workloads in terms of energy efficiency.