Holographic-Type Communication for Digital Twin: A Learning-based Auction Approach

TL;DR

This paper introduces a learning-based auction mechanism for allocating holographic digital twin services, combining holographic communication and digital twin technologies to improve efficiency and user interaction.

Contribution

It proposes a novel auction-based system using deep reinforcement learning to optimize resource allocation in holographic digital twin services.

Findings

Achieves near-optimal social welfare in simulations

Halves auction information exchange cost compared to baseline

Demonstrates effectiveness of learning-based auctioneer

Abstract

Digital Twin (DT) technologies, which aim to build digital replicas of physical entities, are the key to providing efficient, concurrent simulation and analysis of real-world objects. In displaying DTs, Holographic-Type Communication (HTC), which supports the transmission of holographic data such as Light Field (LF), can provide an immersive way for users to interact with Holographic DTs (HDT). However, it is challenging to effectively allocate interactive and resource-intensive HDT services among HDT users and providers. In this paper, we integrate the paradigms of HTC and DT to form a HTC for DT system, design a marketplace for HDT services where HDT users' and providers' prices are evaluated by their valuation functions, and propose an auction-based mechanism to match HDT services using a learning-based Double Dutch Auction (DDA). Specifically, we apply DDA and train an agent acting as the auctioneer to adjust the auction clock dynamically using Deep Reinforcement Learning (DRL), aiming to achieve the best market efficiency. Simulation results demonstrate that the proposed learning-based auctioneer can achieve near-optimal social welfare at halved auction information exchange cost of the baseline method.