MADQN-Enhanced Computation Offloading and Resource Allocation for 6G Low-Altitude Economy Vehicular Networks

Hu, B ORCID: 0000-0003-4821-0448, Liu, H, Du, J ORCID: 0000-0002-0845-4942, López-Benitez, M ORCID: 0000-0003-0526-6687, Wu, C ORCID: 0000-0001-6853-5878, Chu, X ORCID: 0000-0003-1863-6149 and Niyato, D ORCID: 0000-0002-7442-7416 (2026) MADQN-Enhanced Computation Offloading and Resource Allocation for 6G Low-Altitude Economy Vehicular Networks IEEE Transactions on Cognitive Communications and Networking, 12 (99). pp. 2603-2617. ISSN 2332-7731, 2332-7731

Text IEEE_TCCN_2025.pdf - Author Accepted Manuscript Available under License Creative Commons Attribution. Download (9MB) | Preview

Abstract

Air-to-ground communication networks in future sixth-generation (6G) networks are expected to leverage integrated sensing and communication (ISAC) to support the low-altitude economy (LAE). In such networks, a set of unmanned aerial vehicles (UAVs) acting as mobile edge computing (MEC) servers cooperatively process delay-sensitive tasks offloaded by multiple authorised vehicular user equipments (V-UEs). However, the diverse, stringent requirements of ISAC-enabled V-UE services require more intelligent and efficient resource allocation for the LAE-aided vehicle-to-everything (V2X) communications systems. To address this issue, we propose a digital twin (DT)-empowered multi-agent LAE MEC vehicular framework, where the DT technology enables real-time data collection, processing, monitoring, and optimisation in a virtual environment. Meanwhile, each V-UE may offload its delay-sensitive task to a UAV-assisted MEC server. We aim to minimise the long-term average total service delay (which may include the task processing delay and the transmission delay) of all V-UEs, the computation resource allocation at each UAV-assisted MEC server, the transmission power, and the allocation of resource blocks for all V-UEs. To solve the joint optimisation problem, we propose a Multi-agent deep Q-network-based Offloading and Resource allocation Optimisation (MORO) algorithm. Simulation results demonstrate that our proposed algorithm outperforms the benchmarks in terms of the convergence rate and the long-term average total service delay of all V-UEs.

Item Type:	Article
Uncontrolled Keywords:	Servers, Optimization, Vehicle-to-everything, Resource management, Communication networks, Delays, Monitoring, Computational efficiency, Quality of service, Energy consumption, Low-altitude economy, V2X communications, digital twins, UAV, multi-agent deep Q-network, deep reinforcement learning
Divisions:	Faculty of Science & Engineering Faculty of Science & Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User:	Symplectic Admin
Date Deposited:	03 Jul 2025 13:04
Last Modified:	28 Feb 2026 16:38
DOI:	10.1109/TCCN.2025.3586868
Related Websites:	Author Publisher
URI:	https://livrepository.liverpool.ac.uk/id/eprint/3193535
Disclaimer:	The University of Liverpool is not responsible for content contained on other websites from links within repository metadata. Please contact us if you notice anything that appears incorrect or inappropriate.