Joint Caching,Transcoding And Delivery Strategy At Edge For Adaptive 360-Degree Video Streaming

The high-definition and panorama characteristics of 360-degree video enable viewers immersive experience but make its transmission in existing networks more challenging.Specifically,transmitting the entire 360-degree video will consume substantial bandwidth resources and may easily cause network congestion and transmission delay.Actually,a user wearing a head-mounted display can view at any time a partial scene of the entire360-degree video,namely the field of view of the user.Thus,tile-based360-degree video streaming,which combines tiling and adaptive streaming techniques to slice the 360-degree video in time and space,is an effective bandwidth-saving approach.Furthermore,edge caching technology helps alleviating the burden of backhaul links during peak traffic hours.However,The storage capacity of a single server is insufficient for tile-based adaptive360-degree streaming.Thus,instead,jointly employing storage and computing resources of multiple edge servers can better facilitate the adaptive360-degree video streaming.Therefore,this paper firstly proposes to leverage storage and computing resources of multiple edge servers in the edge network to jointly decide on edge caching,transcoding and delivery strategies at the tile-granularity to assist the adaptive 360-degree video streaming.First,this paper designs a transmission framework for the tile-based joint caching,transcoding and delivery scheme at edge and clarifies the functional modules of the edge servers.Based on the proposed transmission framework,this paper mathematically formulates the joint caching,transcoding and delivery decision-making problem as a multivariate nonlinear integer programming problem,aiming at minimizing the aggregate network operational cost and balancing the server load,subject to storage and computation capacity constraints of the edge servers.The formulated optimization problem is NP-hard and intractable.Through decomposition and relaxation approach,the original intractable optimization problem is solved by an effective iterative algorithm.Finally,simulation results show that the proposed iterative algorithm outperforms the existing schemes in terms of resource utilization,cost reduction and server coordination.Previous optimization-based static joint caching,transcoding and delivery strategy at edge is not applicable to the scenario where user requests change dynamically and are difficult to predict accurately.Therefore,this paper proposes another deep-reinforcement-learning-based dynamic joint caching,transcoding and delivery strategy at edge.It aims to dynamically take joint caching,transcoding and delivery actions at edge according to environment states and user requests,to minimize the overall operating costs of the edge network,subjected to the constraints of the network.Specifically,this paper remodel the dynamic transmission model and design the reinforcement learning elements,and build a DDPG-based deep reinforcement learning network to derive the joint decision-making agent.In order to limit the excessive action space in the tile-based adaptive 360-degree video streaming,this paper proposes a caching-decision-alternative-set-based caching actor scheme when implementing the DRL network.Finally,simulation experiments verify that the proposed DRL-based dynamic joint caching,transcoding and delivery strategy at edge can converge fast and achieve near-optimal resource scheduling performance.