Joint Optimization of Data Value and Age of Information in Multi-cluster System with Mixed Data

Age of Information (AoI) is an emerging time-related indicator in the industry. It is often used to evaluate the freshness of received data. Considering a multi-cluster live streaming system with mixed video data and environmental data, a scheduling policy is formulated to jointly optimize the system data value and AoI. To overcome the problem that the effective solution to the optimization problem is difficult to achieve due to the action space being too large, the scheduling policy of the optimization problem is decomposed into two interrelated internal layer and external layer policies. The external layer policy utilizes deep reinforcement learning for channel allocation between clusters. The internal layer policy implements the link selection in the cluster on the basis of the constructed virtual queue. The two-layer policy embeds the internal layer policy of each cluster into the external layer policy for training. Simulation results show that compared with the existing scheduling policy, the proposed scheduling policy can increase the time-averaged data value of received data and reduce the time-averaged AoI.