Collaborative Research: NeTS: Medium: Towards High-Performing LoRa with Embedded Intelligence on the Edge

点击翻译按钮获取中文摘要

LoRa (short for Long Range), a spread-spectrum modulation technique, has emerged in recent years as a promising mechanism to connect billions of low-cost Internet of Things (IoT) devices for wide-area applications such as smart metering, environment monitoring, and logistic tracking. However, current LoRa networks have been observed to have shorter coverage range, lower energy efficiency, and higher deployment cost than originally promised. The fundamental problem is that current LoRa receivers perform poorly when there is complex environmental noise. This project uses the feature extraction capability of modern deep neural networks (DNN) and the computational resources now available on edge devices to create better performing LoRa networks. The success of this project will reduce the cost of deploying and maintaining real-world LoRa networks, and thus will accelerate adoption of wide-area IoT applications which will enhance efficiency of smart cities and other verticals. The project also develops curricular materials for applying machine learning to wireless networking in both undergraduate and graduate programs. This project offers research training opportunities to underrepresented students from diverse groups and age levels. This project designs a new LoRa physical layer to enhance long-distance and low-power LoRa communication. The project includes three parts. (1) Design of a multi-dimension multi-resolution neural-enhanced LoRa decoder that can be used with standard LoRa transmissions in a single-gateway setting. The new decoder improves performance by capturing and processing multi-dimensional features of standard LoRa signals even when the signal strength is far below the noise floor; (2) Co-design of a neural-enhanced encoder-decoder pair for use in a single-gateway setting. The encoder creates a non-standard LoRa transmission that provides a much richer feature space for neural-enhanced decoding and thus further enhances performance in high-noise situations. (3) Co-design a neural-enhanced multi-gateway symbol decoder and a frequency-aware encoder for use in a multi-gateway setting. The design uses the spatial diversity of multiple gateways to enhance the SNR (signal to noise ratio) of the received signals even further. To evaluate the proposed techniques, this project uses hardware-software co-design to develop an end-to-end DNN-empowered LoRa prototype. The code and data generated in the project are available to the research community for further investigation.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.