Deep Reinforcement Learning-Based Task Offloading in 5G-Enabled Multi-Access Edge Computing
Year: 2026
Type: conference-paper
Abstract: The 5G technology offers capabilities such as ultralow latency, high data rates, and reliable communication. This allows for advanced technologies such as smart healthcare services, IoT devices, autonomous vehicles, and data processing. However, extensive dependency on centralized cloud computing causes additional end-to-end delays and congestion. To address this issue, this study proposes implementing an intelligent mechanism for offloading tasks, considering 5G-enabled edge computing and cloud computing. The framework employs a reinforcement learning approach, specifically Q-learning algorithm and Deep Q-Network (DQN) to determine whether tasks should be handled locally or offloaded to the edge layer depending on real-time state of the edge devices and network conditions. This decision-making strategy is quite effective for the minimization of processing latency with much lower energy requirements than the on-device processing approach. In addition to this, the focus on preventing unnecessary offloading of the task results in improved overall network performance. Experimental results demonstrate notable improvements in latency reduction and energy efficiency in 5G edge-cloud networks.
Institution: KLE Technological University
Language: English
Topic: IoT and Edge/Fog Computing
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204800407