A DQN-Based Resource Adaptation Strategy for In-Network Computing-Based Holographic-Type Communication

Holographic-type communication (HTC) provides users with highly immersive and interactive experiences, which require a network with ultralow latency and large bandwidth guarantees. However, it is difficult for traditional networks to support HTC services. To provide immersive experiences, high-quality holograms need to be delivered with low motion-to-photon (MTP) delay. It is related to computing power resources. To address the above problems, we use in-network computing to provide computing resources for HTC service, considering the delay and the perceived quality of holographic content processed by the computing resources. With the joint optimization objectives of load balancing and improving the perceived quality of content, we allocate limited computing resources and bandwidth resources for different service requests and customize the appropriate resource chains. In this work, we employ a deep Q network (DQN) to solve two subproblems—network resource adaptation and optimal route determination—to achieve intelligent and efficient resource adaptation. The simulation results show that the resource adaptation strategy is more effective in terms of the acceptance rate and balancing network resource load.