Abstract : The rapid growth of the internet usage and the distributedcomputing resources of edge devices create a necessity tohave a reasonable controller to ensure efficient utilization of distributedcomputing resources in mobile edge computing (MEC).We envision the future MEC services, where quality of experience(QoE) of the services is further enhanced by software definednetworks (SDNs) capabilities to reduce the application-levelresponse time without service disruptions. SDN, which is not proposedspecifically for edge computing, can in fact serve as an enablerto lower the complexity barriers involved and let the realpotential of edge computing be achieved. In this paper, we investigatethe task offloading and resource allocation problem in wirelessMEC aiming to minimize the delay while saving the battery powerof user device simultaneously. However, it is challenging to obtainan optimal policy in such a dynamic task offloading system. Learningfrom experience plays a vital role in time variant dynamic systemswhere reinforcement learning (RL) takes a long term goal intoconsideration besides immediate reward, which is very importantfor a dynamic environment. A novel software defined edge cloudlet(SDEC) based RL optimization framework is proposed to tacklethe task offloading and resource allocation in wireless MEC. Specifically,Q-learning and cooperative Q-learning based reinforcementlearning schemes are proposed for the intractable problem. Simulationresults show that the proposed scheme achieves 31.39% and62.10% reduction on the sum delay compared to other benchmarkmethods such as traditional Q-learning with a random algorithmand Q-learning with epsilon greedy.
Index terms : Mobile edge computing, resource allocation, software defined cellular networks, task offloading, wireless networks.