DQN for Multi-layer Game Based Mining Competition in VEC Network
Yijing Li, Xuefei Zhang, Huici Wu, Junjie Liu, Dian Tang, Xiaofeng Tao · 2019
Blockchain has been considered as a promising technology to improve the efficiency of data sharing in Vehicular Edge Computing (VEC) by data mining among RSUs and vehicles. In order to obtain more mining reward, RSUs and vehicles will compete to accomplish the data sharing mining task. However, it is easier for RSUs equipped with stronger computing capabilities to win the mining reward. In this condition, some vehicles provide their own computing resources to form a shared resource pool for the competition with RSUs. Meanwhile, the competition among the vehicles sharing the resource pool is non-negligible since some selfish vehicles divide the group reward but without providing resources. In this way, we provide a multilayer game model that involving a deformed N Iterated Prisoners Dilemma (NIPD) among vehicles and a bargain game between resource pool and RSUs. Further, we use multi-agent Deep Q Network (DQN) to achieve the equilibrium between vehicles and RSUs in mining. Finally, numerical results show the optimal strategy can attain a stable data sharing mining system in VEC network.