LLM-Enhanced Deep Reinforcement Learning for Task Offloading in Collaborative Edge Computing

2026-05-07

AI Summary

This work addresses the challenge of achieving low-latency and highly reliable task offloading in collaborative edge computing under uncertainties such as node failures. To this end, we propose LeDRL, a novel framework that, for the first time, integrates a lightweight large language model (LLM) with self-attention-enhanced deep reinforcement learning (DRL) to enable dynamic offloading decisions. LeDRL leverages structured prompting to generate high-level policy priors and incorporates a context alignment module together with a reflective semantic evaluator to facilitate semantic feedback distillation and real-time policy optimization. Experimental results on the CoEdgeSys prototype demonstrate that LeDRL improves task success rates by over 17%, achieves faster convergence, and offers more responsive decision-making. Its efficacy and generalization capability are further validated on resource-constrained edge devices such as NVIDIA Jetson platforms.

Abstract

Collaborative edge computing uses edge nodes in different locations to execute tasks, necessitating dynamic task offloading decisions to maintain low latency and high reliability, especially under unpredictable node failures. Although deep reinforcement learning (DRL) and large language models (LLMs) have shown promise for task offloading, DRL often suffers from high sample inefficiency and local optima, whereas LLMs struggle with real-time decision-making. To address these limitations, we propose \textbf{LeDRL}, a hybrid decision framework that couples a \emph{lightweight LLM} with self-attention-enhanced DRL for real-time task offloading. LeDRL constructs structured, context-aware prompts capturing node status, task semantics, and link dynamics to derive high-level strategy priors. These are selectively processed by a self-attention-based alignment module for context-aware policy optimization. A reflective evaluator distills semantic feedback from past trajectories to guide future prompts, enabling more informative and temporally generalizable LLM queries. Extensive experiments show that LeDRL outperforms baselines in task success rate, convergence speed, and real-time responsiveness across diverse network scales, achieving over 17\% improvement in success rate. Furthermore, we deploy LeDRL on Jetson-based edge devices using our prototype system \textit{CoEdgeSys}, demonstrating its robustness and feasibility under resource constraints. Our code is available at:https://github.com/GalleyG5/LeDRL.git.

Problem

• task offloading
• collaborative edge computing
• node failures
• low latency
• high reliability

Innovation

• LLM-enhanced DRL
• task offloading
• collaborative edge computing
• self-attention alignment
• context-aware prompting