[Paper Review] Stochastic Resource Allocation for Semantic Communication-aided Virtual Transportation Networks in the Metaverse

Wei Chong Ng, Hongyang Du+4 more|arXiv (Cornell University)|Aug 31, 2022

Robotics and Automated Systems19 citations

TL;DR

This paper proposes a stochastic semantic transmission scheme (SSTS) using two-stage stochastic integer programming to minimize virtual transportation network costs in the Metaverse under uncertain user demand. By dynamically allocating reserved and on-demand semantic data subscriptions based on real-world edge device data, SSTS reduces transmission, storage, and energy costs while outperforming EVF and random allocation methods in simulations with real Singapore deployment data.

ABSTRACT

The physical-virtual world synchronization to develop the Metaverse will require a massive transmission and exchange of data. In this paper, we introduce semantic communication for the development of virtual transportation networks in the Metaverse. Leveraging the perception capabilities of edge devices, virtual service providers (VSPs) can subscribe to their preferred edge devices to receive the semantic data of interest. However, the demands of the VSPs are highly dependent on the users that they are serving. To address the resource allocation problem amid stochastic user demand, we propose a stochastic semantic transmission scheme (SSTS) based on two-stage stochastic integer programming. Using real data captured by edge devices we deploy in Singapore, the simulation results show that SSTS can minimize the transmission cost of the VSPs while accounting for the users' demand uncertainties.

Motivation & Objective

• To address the challenge of high and variable data transmission costs in Metaverse virtual transportation networks due to uncertain user demand for semantic data.
• To design a dynamic resource allocation framework that minimizes total operational costs—transmission, storage, and energy—under demand uncertainty.
• To integrate semantic communication with stochastic optimization for AI-driven Metaverse applications, particularly in virtual transportation systems.
• To evaluate the performance of the proposed scheme against baselines like Expected-Value Formulation (EVF) and random allocation under varying on-demand pricing and demand probabilities.

Proposed method

• Formulates a two-stage stochastic integer programming (SIP) model to determine optimal reservation and on-demand subscription plans for virtual service providers (VSPs).
• Uses real semantic data from edge devices deployed in Singapore to model user demand scenarios and similarity scores for data relevance.
• Incorporates recourse actions in the second stage to correct under-subscription when actual demand exceeds reserved capacity.
• Models subscription costs using two-tier pricing: lower-cost reservation bundles and higher-cost on-demand per-transmission fees.
• Solves the SIP model to minimize total cost, balancing fixed reservation fees against variable on-demand costs under probabilistic demand scenarios.
• Employs a scenario-based approach with discrete demand probabilities (e.g., P(λ₁) = 0.2, 0.4) to represent uncertainty in user interest across VSPs.

Figure 1: An illustrative example of the system model with one VSP from a bus company and four edge devices and they are placed at the traffic junction. The VSP is from an autonomous company and has an interest

Experimental results

Research questions

• RQ1 How can semantic communication be leveraged to reduce data transmission costs in Metaverse-based virtual transportation networks?
• RQ2 What is the optimal trade-off between reservation and on-demand subscription plans under uncertain user demand for semantic data?
• RQ3 How does the proposed stochastic integer programming framework adapt to changes in on-demand pricing and demand probabilities?
• RQ4 How does the SSTS scheme compare in cost efficiency to EVF and random allocation strategies under varying demand conditions?

Key findings

• When demand probability P(λ₂) = 1 (i.e., P(λ₁) = 0), both VSPs use only the reservation plan, resulting in the lowest cost, as it is cheaper than on-demand for guaranteed high demand.
• At P(λ₁) = 0.2, VSP 1 switches from reservation to on-demand due to lower required transmissions (200 vs. 300) and higher fixed cost of reservation, while VSP 2 remains on reservation.
• When P(λ₁) ≥ 0.4, both VSPs switch to on-demand, and total cost decreases with increasing P(λ₁), reaching zero when P(λ₁) = 1 (no demand).
• In the varying-interest scenario, VSP 1 selects 11 reservation bundles from smartphone 1 (highest similarity) when P(λ₁) = 0, and increases to 12 bundles as P(λ₁) increases to balance shortfall.
• When P(λ₁) = 1, VSP 1 switches entirely to 10 reservation bundles from smartphone 3, which has the highest similarity score (0.83) in scenario 1.
• The SIP-based SSTS consistently outperforms EVF and random schemes, maintaining stable cost performance under varying on-demand prices, while EVF and random schemes fail to adapt to cost changes.

Figure 2: Locations of the smartphones in Singapore and an illustrative example of a virtual transportation case study.