Online Multi-Task Offloading for Semantic-Aware Edge Computing Systems

TL;DR

The paper tackles bandwidth-limited uplink scenarios in MEC by introducing a semantic-aware, multi-task offloading framework. It defines a semantic extraction factor $\mu_n$, a unified QoE metric that combines latency, energy, and task accuracy, and formulates the joint optimization as an MDP solved via a semantic-aware MAPPO algorithm. Experimental results across text, image, and VQA tasks show QoE gains over semantic-unaware baselines and demonstrate adaptability to user preferences and varying network conditions. The approach offers a practical path to scalable, context-aware edge computing with configurable QoE objectives in 6G-era networks.

Abstract

Mobile edge computing (MEC) provides low-latency offloading solutions for computationally intensive tasks, effectively improving the computing efficiency and battery life of mobile devices. However, for data-intensive tasks or scenarios with limited uplink bandwidth, network congestion might occur due to massive simultaneous offloading nodes, increasing transmission latency and affecting task performance. In this paper, we propose a semantic-aware multi-modal task offloading framework to address the challenges posed by limited uplink bandwidth. By introducing a semantic extraction factor, we balance the relationship among transmission latency, computation energy consumption, and task performance. To measure the offloading performance of multi-modal tasks, we design a unified and fair quality of experience (QoE) metric that includes execution latency, energy consumption, and task performance. Lastly, we formulate the optimization problem as a Markov decision process (MDP) and exploit the multi-agent proximal policy optimization (MAPPO) reinforcement learning algorithm to jointly optimize the semantic extraction factor, communication resources, and computing resources to maximize overall QoE. Experimental results show that the proposed method achieves a reduction in execution latency and energy consumption of 18.1% and 12.9%, respectively compared with the semantic-unaware approach. Moreover, the proposed approach can be easily extended to models with different user preferences.

Figures (16)

• Figure 1: Semantic-aware multi-task offloading system model. Each task is fed into a semantic codec specifically designed for its type.
• Figure 2: Semantic extraction workflow. The semantic extraction factor alters the degree of semantic compression, affecting the computational burden of semantic extraction, the volume of data transmission, and the accuracy of task execution.
• Figure 3: QoE function. The logistic function serves as the scoring function for task accuracy, with its symmetric function scoring execution latency and energy consumption. The offset $x_0$ represents the latency, energy consumption, or task accuracy of local execution.
• Figure 4: The framework of the MAPPO algorithm.
• Figure 5: The impact of $A$ on $J^{CLIP}$.