An application specific vector processor for CNN-based massive MIMO positioning

Research output: Chapter in Book/Report/Conference proceedingPaper in conference proceedingpeer-review


This paper sets out to create an implementation for fingerprint-based positioning using massive multiple-input multiple-output (MIMO) technology, by means of deep convolutional neural networks (CNN), and utilizing the wireless channel state information (CSI). Due to the sheer volume of computational requirements imposed by CNN processing, an accelerator-assisted design is well-suited to the task at hand. Consequently, an application specific instruction set processor (ASIP) is designed to combine flexibility with implementation efficiency. This ASIP is equipped with vector processing capabilities employing a single instruction multiple data (SIMD) scheme, and additionally has a very large instruction word (VLIW) architecture to further exploit instruction-level parallelism. A configurable 2D array of processing engines (PE) is integrated into the processor, in a tightly coupled manner, to accelerate the CNN operation. Synthesis results will be demonstrated using the GF-22 nm FD-SOI technology with a clock frequency of 555 MHz. The system can achieve a throughput of 271 positionings/s, with an average positioning error of 3.5 λ (40 cm) at a carrier frequency of 2.6 GHz.

Original languageEnglish
Title of host publication2021 IEEE International Symposium on Circuits and Systems, ISCAS 2021 - Proceedings
PublisherIEEE - Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728192017
Publication statusPublished - 2021
Event53rd IEEE International Symposium on Circuits and Systems, ISCAS 2021 - Daegu, Korea, Republic of
Duration: 2021 May 222021 May 28

Publication series

NameProceedings - IEEE International Symposium on Circuits and Systems
ISSN (Print)0271-4310


Conference53rd IEEE International Symposium on Circuits and Systems, ISCAS 2021
Country/TerritoryKorea, Republic of

Bibliographical note

Funding Information:
The design has been synthesized using the GF-22 nm FD-X technology. Table I shows the area breakdown occupied by different modules in the system. The whole system takes up a cell area of around 1 mm2, with an operating frequency of 555 MHz. The vector memory takes the bulk of the area which amounts to around 75% usage. This global SRAM serves as a buffer for filter weights, input features, and intermediate tensors. The programmable portion of the processor requires nearly the same area as the specialized CNN engine, together adding up to almost 20% of the area budget. The power consumption of the architecture is estimated at 150 mW (using power annotated simulation results), which does not account for off-chip (e.g. DRAM) accesses. V. CONCLUSION In the post-Moore era the focus has shifted from component miniaturization to algorithms/software performance-engineering along with specialization of computer architecture. In this paper we investigated the utilization of an ASIP vector processor, allied with a dedicated CNN engine, to implement user positioning in a Massive MIMO-based wireless system. The processor achieves a frequency of 555 MHz, and can churn out 271 user positions per second, with an average distance error of 3.5 λ, and a power draw of 150 mW. VI. ACKNOWLEDGMENTS This work is supported by Ericsson’s Massive MIMO project. The authors would also like to thank Synopsys for providing their tool ASIP Designer.

Publisher Copyright:
© 2021 IEEE

Copyright 2021 Elsevier B.V., All rights reserved.

Subject classification (UKÄ)

  • Computer Engineering


Dive into the research topics of 'An application specific vector processor for CNN-based massive MIMO positioning'. Together they form a unique fingerprint.

Cite this