Multi-leader single-follower stackelberg game task offloading and resource allocation based on selection optimization in Internet of Vehicles

• Published in: IEEE access Vol. 11; p. 1
• Main Authors: Li, Yanqiang, Li, Lijuan, Xia, Yang, Zhang, Daifeng, Wang, Yong
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Computation offloading; Costs; Delay; Delays; Edge computing; Game theory; Games; Internet of Vehicles; Mobile computing; multi-objective optimization; Multiple objective analysis; Network latency; Optimization; Resource allocation; Servers; Stackelberg game; task offloading; Vehicular dynamics; Vehicular networks
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2023.3280412

Mobile edge computing (MEC) provides a new solution to meet the latency-sensitive and computation-intensive application requirements of vehicle networks. Different from existing work, a novel multi-leader single-follower game model is designed, in which each leader can both compete and cooperate with each other. In the case of the leader, a profit maximization optimization problem is proposed. Since the problem is a convex function, there is a Nash equilibrium of the game. Each leader sets the optimal unit price by the amount of computational resources required by the follower. For the followers, a multi-objective optimization problem is formulated with the objective of minimizing the task processing delay and cost. Then, the task offloading and resource allocation based on selection optimization(TRSO) algorithm is proposed to achieve a tradeoff between latency and cost. Specifically, the computational resources are fixed and the Karush-Kuhn-Tucker (KKT) algorithm is used to jointly optimize the task division ratio and the bandwidth allocation ratio to minimize the task processing delay. In addition, the dichotomous method is used to optimize the computational resources of edge servers (ESs) under the task latency constraint so that the total cost is minimized. Simulation results confirm that the proposed TRSO is superior to the task offloading and resource allocation optimization (TRO) algorithm. Under the same delay, compared with TRO, the task processing cost of TRSO is reduced by approximately 83.3%, significantly reducing the overall task processing cost.