.png){kind=icon}
.png){kind=icon}
Return
Multi-task classification offloading strategy based on reinforcement learning for mobile edge computing
.png){kind=icon}
PRE
AI
DOI：10.1007/s11227-025-07693-x.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
To address the resource competition issue caused by multi-task concurrent computation offloading in the Internet of Vehicles (IoV), this paper proposes the MCO-DQN strategy, which collaborates with vehicles and edge servers to establish a multi-task concurrent classification offloading model. The model constructs a constrained multi-objective optimization problem with latency and energy consumption as dual objectives, aiming to achieve dual optimization and dynamic balance between both performance metrics. Based on the task characteristics, different state coefficient weights are assigned, and computation offloading is formalized as a Markov Decision Process to meet the classification offloading requirements of different types of tasks. By combining real-time task status with high-dimensional state modeling and the adaptive capabilities of the DQN algorithm, the global optimal offloading decision is determined, accurately and efficiently implementing multi-task classification offloading in the dynamic and complex scenarios of IoV. Experimental results show that the proposed strategy demonstrates superior performance in terms of latency, energy consumption, success rate, and convergence, making it a practical and optimal solution for multi-task classification offloading in IoV.
Keywords:
Multi-task concurrency
Classification offloading
Dynamic decision
Multi-objective optimization
Journal
IF:
2.7
Papers:
990
Citations:
1.0W