Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Coded Multi-User Information Retrieval with a Multi-Antenna Helper Node

Milad Abolpour, MohammadJavad Salehi, Soheil Mohajer, Seyed Pooya Shariatpanahi, Antti Tölli

TL;DR

The paper addresses multi-user information retrieval with partial data access in a wireless network aided by a multi-antenna base station. It introduces a coded caching–inspired UL-DL strategy that transforms uplink transmissions into DL codewords, achieving a unified NDT of $T_{\rm UL}=T_{\rm DL}=\frac{K-t}{t+L}$ and linking to the optimal DL MISO-CC performance. The main contribution is a joint UL-DL transmission design that uses over-the-air computation and zero-forcing beamforming to create and deliver CC-like codewords, enabling each user to recover its desired content from cached and received data. The results have practical impact for industrial IoT, distributed coded computing, and cache-enabled wireless networks by enabling low-latency, scalable data exchange even when the helper BS does not have direct library access.

Abstract

A novel coding design is proposed to enhance information retrieval in a wireless network of users with partial access to the data, in the sense of observation, measurement, computation, or storage. Information exchange in the network is assisted by a multi-antenna base station (BS), with no direct access to the data. Accordingly, the missing parts of data are exchanged among users through an uplink (UL) step followed by a downlink (DL) step. In this paper, new coding strategies, inspired by coded caching (CC) techniques, are devised to enhance both UL and DL steps. In the UL step, users transmit encoded and properly combined parts of their accessible data to the BS. Then, during the DL step, the BS carries out the required processing on its received signals and forwards a proper combination of the resulting signal terms back to the users, enabling each user to retrieve the desired information. Using the devised coded data retrieval strategy, the data exchange in both UL and DL steps requires the same communication delay, measured by normalized delivery time (NDT). Furthermore, the NDT of the UL/DL step is shown to coincide with the optimal NDT of the original DL multi-input single-output CC scheme, in which the BS is connected to a centralized data library.

Coded Multi-User Information Retrieval with a Multi-Antenna Helper Node

TL;DR

The paper addresses multi-user information retrieval with partial data access in a wireless network aided by a multi-antenna base station. It introduces a coded caching–inspired UL-DL strategy that transforms uplink transmissions into DL codewords, achieving a unified NDT of and linking to the optimal DL MISO-CC performance. The main contribution is a joint UL-DL transmission design that uses over-the-air computation and zero-forcing beamforming to create and deliver CC-like codewords, enabling each user to recover its desired content from cached and received data. The results have practical impact for industrial IoT, distributed coded computing, and cache-enabled wireless networks by enabling low-latency, scalable data exchange even when the helper BS does not have direct library access.

Abstract

A novel coding design is proposed to enhance information retrieval in a wireless network of users with partial access to the data, in the sense of observation, measurement, computation, or storage. Information exchange in the network is assisted by a multi-antenna base station (BS), with no direct access to the data. Accordingly, the missing parts of data are exchanged among users through an uplink (UL) step followed by a downlink (DL) step. In this paper, new coding strategies, inspired by coded caching (CC) techniques, are devised to enhance both UL and DL steps. In the UL step, users transmit encoded and properly combined parts of their accessible data to the BS. Then, during the DL step, the BS carries out the required processing on its received signals and forwards a proper combination of the resulting signal terms back to the users, enabling each user to retrieve the desired information. Using the devised coded data retrieval strategy, the data exchange in both UL and DL steps requires the same communication delay, measured by normalized delivery time (NDT). Furthermore, the NDT of the UL/DL step is shown to coincide with the optimal NDT of the original DL multi-input single-output CC scheme, in which the BS is connected to a centralized data library.
Paper Structure (14 sections, 65 equations, 2 figures, 3 tables)

This paper contains 14 sections, 65 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. Reference Strategies
  4. Strategy A
  5. Strategy B
  6. The New UL-DL Transmission Strategy
  7. An Illustrative Example
  8. UL Step
  9. DL Step
  10. Conclusion
  11. Proof of Lemma \ref{['lemma: power allocation A']}
  12. Proof of Theorem \ref{['theorem: load UL A']}
  13. Proof of Theorem \ref{['theorem: DL']}
  14. An Example for $K=5$, $t=2$ and $L=3$

Figures (2)

  • Figure 1: System model for a coded data retrieval network with $K$ single antenna users and a BS with $L$ antennas: A $\gamma$ portion of the entire library is either locally generated or stored beforehand in the cache memories of users.
  • Figure 2: Comparison between the achievable UL normalized delivery time of the proposed scheme with strategies A and B in a MISO network with $K=10$ and $L=4$.