| The development of communication technology has provided the basic conditions for the popularity of video services,and the popularity has driven the explosive growth of video traffic in recent years.However,the huge video traffic puts heavy communication and computing pressure on mobile edge networks,making the centralized data processing model suffer a serious crisis.First,due to the wide distribution of mobile user devices and the ”long tail” of user video access(uneven content popularity),there is a large amount of redundant video content in the mobile edge network,increasing the transmission burden of the mobile edge network.Second,the sensing video upload in the mobile edge network is affected by wireless connections,network topology,equipment load in the mobile edge network.For this,how to ensure fast and efficient perceptual video upload is one of the current hot research areas.Meanwhile,the time-varying wireless connections and moving sensing devices also make it very difficult to provide users with low-latency,highquality,and highly adaptable video services in the mobile edge network.The approach of deploying video caching on user-side devices provides a promising video distribution mechanism.However,the mobile edge caching-based content delivery needs to overcome the dependence on the core network and wireless communication infrastructure,enhance the adaptability of content delivery,and provide the most suitable video quality for the user terminals.Therefore,the research on the video transmission and caching mechanisms in mobile edge networks has practical and theoretical significance,but it also faces many technical difficulties.Aiming at the mentioned-above problems,combined with the latest research results,this work has made contributions in the following three aspects:1.This thesis investigates the problem of online extracting key content from video streams at the resource-and capability-constrained mobile user devices in mobile edge networks,so that the receiving user can understand the main information of the video before accessing the full video content and the unnecessary video transmissions can be reduced.This thesis designs a low-complexity online video summarization algorithm,adapting to the diversity of user requirements and the dynamics of the mobile edge network by extracting different combinations of representative video segments.This thesis defines the video segment coverage as well as its left and right anchors.By minimizing the video summary error and automatically detecting abnormal values,the ”elastic”video summary problem in the latency-aware mobile edge network is formulated as a 0-1matrix optimization problem.In order to solve this NP-hard complex sparse optimization problem,this thesis proposes an Elastic Video Summarization(EVS)algorithm to dynamically adjust the number of selected representative video segments.In order to reduce the algorithm complexity,this thesis makes a theoretical analysis on the essential attributes of the proposed algorithm,and then add the anchor search and the representative search algorithms into EVS based on the analytical results to substantially narrow the search scope.The validation results demonstrate that the proposed algorithm has the strong network and user adaptability,and can significantly improve bandwidth utilization and reduce time consumption.2.This thesis conducts research on the mobile crowd sensing video upload problem,utilizes the device-to-device(D2D)communication technology to deal with the poor signal conditions,and integrates the content-based dynamic routing with the in-network de-duplication and de-redundancy technologies to reduce the transmitted video sizes.By defining the ”production-storage-transmission” paradigm,the video processing and forwarding strategies are proposed.Then this thesis characterizes the D2 D collaborative video upload problem with two rounds of de-duplication and de-redundancy operations as a multi-stage decision-making problem.Since the decision-making problem involves the uncertain moving transmission destinations in the mobile edge network,the thesis designs a Mutual-Assisted Video Upload Algorithm to convert the decision-making problem into the combination of a knapsack problem and a simple linear programming problem.The numerical results show that our work significantly reduces the video transmission latency and effectively improves the load balance among mobile edge devices.3.To accommodate the dynamic mobile edge networks,aiming at the Scalable Video Coding(SVC)videos,this thesis investigates the problem of collaborative cache placement for SVC video layers in mobile edge networks with D2 D support.From the user's perspective,this thesis jointly considers user perceived latency,cache storage cost,and caching load,and exploits the dependency constraints among SVC video layers to formulate the cache placement problem of SVC video layers as a 0-1 integer programming problem with multi-source and single-source placement support.Also,the thesis designs a greedy Layer-Wise Load-Aware Greedy Cache Placement Algorithm based on a submodularity analysis,in which the requested SVC video layers are classified as the shared layers with multiple source placement and the non-shared layers with single source placement.The shared layer placement problem is transformed into a single-source capacity facility location problem and the non-shared layer placement problem is converted to a simple linear programming problem.The validation results demonstrate that the proposed algorithm can achieve the lowest total cache cost and effectively improve time efficiency and load balance among caches. |
PDF Full Text Request