Reinforcement Learning Based Prioritized Cross-Area Resource Management for Vehicular Cloud Networks

It is anticipated that cross-area resource allocation for Vehicular Cloud Networks (VCNs) which combine Internet of Vehicles with cloud computing becomes an issue for research. In this paper, we propose a prioritized resource allocation scheme specifically in cross-area scenario which is confronted with shortage of resources and featured by high dynamic. And we consider two categories of service requests: local requests and migrated requests, the former have priority to use the computation resources to guarantee local users satisfaction. A reinforcement learning approach with hotbooting technique is applied to determine the best-performing actions through cooperation of RSUs and remote cloud pursuant to the optimized target of maximizing overall system expected reward. Numerical results are presented to show the performance of our proposed strategy in terms of action probability as well as system reward.