Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Uplink Rate-Splitting Multiple Access for Mobile Edge Computing with Short-Packet Communications

Jiawei Xu, Yumeng Zhang, Yunnuo Xu, Bruno Clerckx

TL;DR

This work addresses latency-sensitive MEC with two users offloading tasks to an MEC server under short-packet communications. It proposes an uplink RSMA framework that splits one user’s task into two streams while the other user remains unsplit, and analyzes the overall SCP under finite-blocklength constraints. An alternating-optimization algorithm jointly optimizes offloading factors, power allocation across RSMA streams, and rate-splitting parameters, combining convex subproblems (SOCP) and Taylor approximations to handle non-convex coupling. Numerical results show that RSMA outperforms NOMA in SCP and enables lower latency, highlighting its practical potential for reliable, low-latency MEC in wireless networks with stringent blocklength and reliability requirements.

Abstract

In this paper, a Rate-Splitting Multiple Access (RSMA) scheme is proposed to assist a Mobile Edge Computing (MEC) system where local computation tasks from two users are offloaded to the MEC server, facilitated by uplink RSMA for processing. The efficiency of the MEC service is hence primarily influenced by the RSMA-aided task offloading phase and the subsequent task computation phase, where reliable and low-latency communication is required. For this practical consideration, short-packet communication in the Finite Blocklength (FBL) regime is introduced. In this context, we propose a novel uplink RSMA-aided MEC framework and derive the overall Successful Computation Probability (SCP) with FBL consideration. To maximize the SCP of our proposed RSMA-aided MEC, we strategically optimize: (1) the task offloading factor which determines the number of tasks to be offloaded and processed by the MEC server; (2) the transmit power allocation between different RSMA streams; and (3) the task-splitting factor which decides how many tasks are allocated to splitting streams, while adhering to FBL constraints. To address the strong coupling between these variables in the SCP expression, we apply the Alternative Optimization method, which formulates tractable subproblems to optimize each variable iteratively. The resultant non-convex subproblems are then tackled by Successive Convex Approximation. Numerical results demonstrate that applying uplink RSMA in the MEC system with FBL constraints can not only improve the SCP performance but also provide lower latency in comparison to conventional transmission scheme such as Non-orthogonal Multiple Access (NOMA).

Uplink Rate-Splitting Multiple Access for Mobile Edge Computing with Short-Packet Communications

TL;DR

This work addresses latency-sensitive MEC with two users offloading tasks to an MEC server under short-packet communications. It proposes an uplink RSMA framework that splits one user’s task into two streams while the other user remains unsplit, and analyzes the overall SCP under finite-blocklength constraints. An alternating-optimization algorithm jointly optimizes offloading factors, power allocation across RSMA streams, and rate-splitting parameters, combining convex subproblems (SOCP) and Taylor approximations to handle non-convex coupling. Numerical results show that RSMA outperforms NOMA in SCP and enables lower latency, highlighting its practical potential for reliable, low-latency MEC in wireless networks with stringent blocklength and reliability requirements.

Abstract

In this paper, a Rate-Splitting Multiple Access (RSMA) scheme is proposed to assist a Mobile Edge Computing (MEC) system where local computation tasks from two users are offloaded to the MEC server, facilitated by uplink RSMA for processing. The efficiency of the MEC service is hence primarily influenced by the RSMA-aided task offloading phase and the subsequent task computation phase, where reliable and low-latency communication is required. For this practical consideration, short-packet communication in the Finite Blocklength (FBL) regime is introduced. In this context, we propose a novel uplink RSMA-aided MEC framework and derive the overall Successful Computation Probability (SCP) with FBL consideration. To maximize the SCP of our proposed RSMA-aided MEC, we strategically optimize: (1) the task offloading factor which determines the number of tasks to be offloaded and processed by the MEC server; (2) the transmit power allocation between different RSMA streams; and (3) the task-splitting factor which decides how many tasks are allocated to splitting streams, while adhering to FBL constraints. To address the strong coupling between these variables in the SCP expression, we apply the Alternative Optimization method, which formulates tractable subproblems to optimize each variable iteratively. The resultant non-convex subproblems are then tackled by Successive Convex Approximation. Numerical results demonstrate that applying uplink RSMA in the MEC system with FBL constraints can not only improve the SCP performance but also provide lower latency in comparison to conventional transmission scheme such as Non-orthogonal Multiple Access (NOMA).

Paper Structure

This paper contains 22 sections, 47 equations, 4 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Contributions
  3. Organization
  4. Notations
  5. System Model
  6. Tasks Offloading Phase
  7. RSMA
  8. NOMA
  9. Tasks Computation Phase
  10. Successful Computation Probability Analysis
  11. Successful Offloading Probability
  12. Successful Execution Probability
  13. Problem Formulation and Algorithm
  14. Offloading factor Optimization
  15. Transmit Power Optimization
  16. ...and 7 more sections

Figures (4)

  • Figure 1: The -aided system with two users.
  • Figure 2: The successful computation probability performance of RSMA and NOMA versus the size of tasks with two different transmit SNR averaged over 100 random channel realisations. $M_{2}=5.5$k bits. (a) N=250; (b) N=500; (c) N=750; (c) N=1000.
  • Figure 3: The successful computation probability performance of RSMA and NOMA versus blocklength with two different task sizes averaged over 100 random channel realisations. Transmit SNR is 15 dB.
  • Figure 4: The successful computation probability performance of RSMA and NOMA versus transmit SNR with four different finite blocklength averaged over 100 random channel realisations. $M_{1}=7$k bits. $M_{2}=7$k bits.

Theorems & Definitions (1)

  • Remark 1