Power-Aware Platform Design for Wireless Sensor Networks

Опубликовано 6 сентября 2016, 17:08

Designing wireless sensor networks that can provide meaningful services in every-day life applications requires the ability to continuously monitor humans and understand what they do over space and time. However, monitoring humans is a very complex task that might involve a set of heterogeneous sensors ranging from simple motion or RFID sensors to more complex camera sensors. Typical low-power microprocessors cannot be used to concurrently acquire, process and communicate the information produced from all of these sensing modalities. As a result, a shift from low-power to power-aware sensor node platform design is needed, where the required application processing and communication capabilities must become available at the minimum possible power overhead. In this design paradigm more capable processors or even multiple heterogeneous processors and/or radios might be used. Networked embedded systems with heterogeneous processors and/or radios extend the energy/timing trade-off flexibility and provide the opportunity to fine tune resource utilization. In this talk, I will present specific power-aware design techniques that can be employed to achieve energy-efficient task scheduling strategies and show how several design bottlenecks and tradeoffs can be exploited to improve the performance and accuracy of these strategies. These design techniques, bottlenecks and tradeoffs will be presented in the context of two different prototype power-aware platforms I have designed and implemented, the XYZ sensor node and the mPlatform architecture.