Fig. 3
From: Dynamic handoff policy for RAN slicing by exploiting deep reinforcement learning
DQN-based handoff process. As shown in the figure, the UE periodically measures and reports the obtained QoS to the source BS or NS, and the source SDN controller checks if the handover trigger condition is satisfied. Then the UE uses DQN to select the target BS and NS and sends the handover request to the corresponding SDN controllers. After the confirmation of handover decision, this handover is executed by the target and the source SDN controllers. Before the handover is completed, the target SDN controller calculates the reward value of this handover decision and then broadcasts the reward to the UEs. The UEs served by the same type of the target NS use DQN to update Q-table. Finally, the resource of source BS and NS is released by the SDN controller