Research On Coded Caching Techniques For Video Transmission

With the arrival of the big data era and the development of mobile terminals,the load of conventional communication networks is increasing,while video traffic is the impetus.Caching is a promising approach to deal with the increasing video traffic load,and coded caching is a new technique to alleviate the communication load by utilizing the multi-cast opportunities created by the contents in the caches,which is the focus of this thesis:Firstly,in order to investigate the problem that,the packet number of each file has to be very large,we introduced a new concept of placement delivery array(PDA),to depict both placement and delivery phases in a single array.PDA is able to depict the scheme proposed by Maddah-Ali and Niesen,and it is proved that,the Ali-Niesen scheme achieves the best coding gain with the least possible packet number among the so called regular PDAs.We also propose two new classes of PDAs.Compared to the Ali-Niesen scheme,the new PDAs save an exponentially increasing factor with the number of users in packet number,while the loss in coding gain is only one.Moreover,we established the connection between PDA and the strong edge colouring of bipartite graphs,through which we further found out a new class of PDA.Secondly,for the decentralized coded caching schemes,we present a lower bound for the average rate of a class of placement and delivery schemes.Starting from the lower bound,we further propose a new placement scheme and two new delivery schemes,i.e.,SGD and PGD.SGD searches the cooperative opportunities extensively,thus it achieves the best performance but with high complexity when the number of users is large.PGD can be implemented with low complexity in this case,and achieve the similar performance with SGD.Compared to the known algorithms,the new delivery schemes are able to utilize the multi-cast opportunities created by the caches more sufficiently.Simulation results indicate that,the gain of the new schemes is relatively large,especially in the case of large number of users and limited packet number.Thirdly,in the case that the packet number can be arbitrarily large,we prove that the performance ratio of the decentralized and centralized coded caching schemes proposed by Maddah-Ali and Niesen is between 1 and 1.5.Furthermore,both the upper bound and the lower bound are achievable.The upper bound tightens the known bound from 12 to 1.5.This illustrates that,in case of non-limited packet number,the decentralized algorithm can achieve similar performance with the centralized algorithm.Furthermore,when the number of users goes to infinity,the ratio will approach 1.Finally,for centralized caching schemes,we generalized the concept of PDA to D2D networks,and give the concept of placement delivery array that is applicable to D2D networks(i.e.,DPDA).We establish a general method of constructing regular DPDAs from regular PDAs.Thus,all such PDAs are applicable to D2D networks.For the decentralized algorithm,due to its flexibility,we apply it into a network where the users have access to the network randomly and the caches need to be updated online.Our analysis shows that,compared to the offline scenario,the average extra rate needed to update the caches is no larger than the arrival probability of file library at the server.