Introduction

This repository contains the source code for paper "RAScatter: Achieving Energy-Efficient Backscatter Readers via AI-Assisted Power Adaptation (IoTDI'22)". Check our paper for details. Technically, this work makes three possibly meaningful attempts:

Multi-Rate Computational RFID Tag: It modifies the firmware of WISP-5.1 Tag (160kHz) to support backscattering with FM0/M2/M4/M8 encodings, while the original WISP-5.1 Tag only supports FM0. This makes WISP possible with rate adaptation schemes. See our core assembly code to generate different encodings here.
AI for Backscatter: A neural network based predictor is used at runtime to (a) adjust reader's Tx power and (b) let the reader send commands to the tag to adjust its backscatter data rate, i.e., switch encoding schemes. Such adaptation aims to achieve the user-specified backscatter goodput with minimally required RF power from the reader. To implement this, we managed to jointly compile the GNU Radio code with TensorFlow v1 C API.
Knowledge-guided AI design: The neural network design is guided by backscatter theory. It adopts a modular structure and adds a special regularizer (Section IV.C in paper) to training for better generazability.

😌 Sadly, our efforts were paid towards a wrong direction: Most people think it's unnecessary to save reader's RF power since they are usually plugged into the wall, though it could benefit several portable readers.

☺️ The code is less likely to be reused. The main purpose of this code releasing is to tell we indeed built an end-to-end system where AI and backscatter are magically integrated together, though in a less recognized way. The tag's code looks fine, but the reader's and NN's training code looks messy. I may find sometime to clean them up in the future.

Citation

🤗 If you find our efforts inspired your work, please kindly cite our paper:

@inproceedings{huang2022rascatter,
  title={RAScatter: Achieving Energy-Efficient Backscatter Readers via AI-Assisted Power Adaptation},
  author={Huang, Kai and Chen, Ruirong and Gao, Wei},
  booktitle={2022 IEEE/ACM Seventh International Conference on Internet-of-Things Design and Implementation (IoTDI)},
  pages={1--13},
  year={2022},
  organization={IEEE}
}

pittisl/RAScatter

Introduction

Citation