Entropy-Aware Task Offloading in Mobile Edge Computing

Mohsen Sahraei Ardakani; Hong Wan; Rui Song

cs.NIMar 16, 2026

Entropy-Aware Task Offloading in Mobile Edge Computing

Authors: Mohsen Sahraei Ardakani, Hong Wan, Rui Song

Abstract

Mobile Edge Computing (MEC) technology has been introduced to enable could computing at the edge of the network in order to help resource limited mobile devices with time sensitive data processing tasks. In this paradigm, mobile devices can offload their computationally heavy tasks to more efficient nearby MEC servers via wireless communication. Consequently, the main focus of researches on the subject has been on development of efficient offloading schemes, leaving the privacy of mobile user out. While the Blockchain technology is used as the trust mechanism for secured sharing of the data, the privacy issues induced from wireless communication, namely, usage pattern and location privacy are the centerpiece of this work. The effects of these privacy concerns on the task offloading Markov Decision Process (MDP) is addressed and the MDP is solved using a Deep Recurrent Q-Netwrok (DRQN). The Numerical simulations are presented to show the effectiveness of the proposed method.

Figures & tables

Symbol	Description
$d_{n}$	number of new tasks at timestep n
$b_{n}$	number of tasks in the buffer at timestep n
$g_{n}$	channel quality at timestep n
$q_{n}$	number of buffered tasks at timestep n
$t_{n}$	number of offloaded tasks at timestep n
$l_{n}$	number of tasks processed locally at timestep n

Table 1: List of notations

Panel of Figure 1 — Figure 1: Comparison of neural network models used in DQN and DRQN

Figure 2: Learning Curves of the trained agents

Parameter	Value
Learning episodes	$10^{3}$
Discount Factor $\gamma$	$0.9$
Exploration rate $\epsilon$	$1\text{ with }0.995$ decay
Recurrent Q-Network layers	$6$
RNN layer Type (hidden units)	3 $\times$ GRU(128), 2 $\times$ FC(128), FC( $\|\mathcal{A}\|$ )
Replay buffer size	$10^{5}$

Table 2: DQRN parameters

Panel of Figure 3 — Figure 3: Comparison of reward macros for the agents

Panel of Figure 4 — Figure 4: Comparison of privacy metrics for the agents

Agent	$\bar{C}$	$\bar{L}$ (s/task)	$\bar{E}$ (J/task)	H(D,T)	H(G,T)	privacy aware metric
Entropy aware DQN	1.663	0.163	1.499	3.258	2.019	2.872
Entropy aware DRQN(128)	1.95	0.411	1.539	4.481	3.335	3.713
Entropy aware DRQN(160)	1.878	0.338	1.54	4.502	3.338	3.28
None Private	0.861	0.112	0.748	2.623	1.818	0.004
Pravacy Aware DQN (He et. al)	1.273	0.104	1.169	2.405	1.373	7.489

Table 3: Comparison of performances of the trained models and the baselines

Panel of Figure 5 — Figure 5: Privacy reward weight matched to $\theta$

$\lambda$	$\bar{C}$	$\bar{L}$ (s/task)	$\bar{E}$ (J/task)	H(D,T)	H(G,T)	Privacy aware Metric
2	1.538	0.601	0.938	3.649	2.574	0.536
5	1.596	0.392	1.204	4.114	2.98	1.581
8	1.709	0.299	1.41	4.318	3.202	2.364
10	1.95	0.411	1.539	4.481	3.335	3.713
16	2.275	0.422	1.853	4.675	3.513	6.474
20	2.454	0.418	2.036	4.734	3.598	8.4

Table 4: Comparison of performances of the trained DRQN128 models with different privacy reward weights

Edge Computing

Edge computing places computation and AI near data sources to reduce latency, bandwidth use, and reliance on centralized clouds.

See all 198 papers

Multi-Turn Reasoning LLMs for Task Offloading in Mobile Edge Computing
Apr 8, 2026Ning Yang, Chuangxin Cheng, Haijun ZhangEdge Computing LLM Reasoning
Institute of Automation, Chinese Academy of Sciences, Beijing, 100190, China · School of Mechano-Electronic Engineering, Xidian University, Xi’an, 710071, China · School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing, 100083, China
QAROO: AI-Driven Online Task Offloading for Energy-Efficient and Sustainable MEC Networks
Apr 28, 2026Yongtao Yao, Yao Yang, Haorui Shi +3Edge Computing Wireless Resource Allocation
Heterogeneous Tasks Offloading in Vehicular Edge Computing: A Federated Meta Deep Reinforcement Learning Approach
May 18, 2026Yaorong Huang, Jingtao Luo, Xuechao WangFederated RL Edge Computing
∗The Hong Kong University of Science and Technology (Guangzhou) · †Chengdu Neusoft University

Apr 8, 2026J/K move · Enter open · S save

Multi-Turn Reasoning LLMs for Task Offloading in Mobile Edge Computing

Ning Yang, Chuangxin Cheng, Haijun Zhang

Institute of Automation, Chinese Academy of Sciences, Beijing, 100190, China · School of Mechano-Electronic Engineering, Xidian University, Xi’an, 710071, China · School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing, 100083, China

Edge Computing LLM Reasoning RL for LLM Agents

Open paper AskPDF arXiv

Emerging computation-intensive applications impose stringent latency requirements on resource-constrained mobile devices. Mobile Edge Computing (MEC) addresses this challenge through task offloading. However, designing effective policies remains difficult due to dynamic task arrivals, time-varying channels, and the spatio-temporal coupling of server queues. Conventional heuristics lack adaptability, while Deep Reinforcement Learning (DRL) suffers from limited generalization and architectural rigidity, requiring retraining when network topology changes. Although Large Language Models (LLMs) offer semantic reasoning capabilities, standard Supervised Fine-Tuning (SFT) yields myopic policies that greedily minimize immediate latency without accounting for long-term system evolution. To address these limitations, we propose COMLLM, a generative framework that enables foresighted decision-making in MEC systems. COMLLM integrates Group Relative Policy Optimization (GRPO) with a Look-Ahead Collaborative Simulation (LACS) mechanism, which performs multi-step Monte Carlo rollouts while jointly modeling server queue dynamics. By incorporating these rollouts into the reward design, the framework captures the long-term impact of current decisions on future system states. Experimental results demonstrate that COMLLM achieves near-optimal latency and improved load-balancing fairness. Notably, it exhibits zero-shot topological scalability, allowing a model trained on small-scale networks to generalize to larger, unseen topologies without retraining, outperforming SFT, DRL, and heuristic baselines.

Abstract

Edge Computing

Explore similar work

Multi-Turn Reasoning LLMs for Task Offloading in Mobile Edge Computing

QAROO: AI-Driven Online Task Offloading for Energy-Efficient and Sustainable MEC Networks

Heterogeneous Tasks Offloading in Vehicular Edge Computing: A Federated Meta Deep Reinforcement Learning Approach