With the rapid development of network communications, the available bandwidth becomes more and more insufficient. Therefore, how to utilize the network resource efficiently becomes a very important research topic in modern network communication. With simple coding scheme and achieveing max-flow bound of network multicast in muticast communication, linear network coding has been widely regarded as an effective technique and the muticast capacity can be achieved through regarding a block of data as a vector over a certain base field and allowwing a node to apply a linear transformation to a vector before passing it on, which can not always be achieved by traditional replicating and routing. Furthemore, through network coding, it can be obtained the benefits of saving in banwidth, load balancing, enhancing the robustness and heightening the efficiency of error correction and so on.In this thesis, it is investigated that the theory of network coding and the construction of linear network coding in muticast network.The integration mechanism of Q-Learning with linear network coding is lucubrated and a random multi-layer network model is explorered to verify the performance of the proposed scheme.The core contents of the work are the following points.(1) Deterministic polynomial time algorithms and distributed random network coding for one-source multicast linear network code construction is researched and it is presented a simple random network coding algorithm with polynomial time complex that achieves coding construction by chosing coding coefficient randomly at each node.(2) In one-source muticast network, linear network coding based on reinforcement learning is lucubrated and it is presented a Q-Learning linear network coding algorithm that applying roulette technique at each node and permitting sink nodes to assign corresponding reward to the incoming edges through checking the incoming codes,which can achieve network coding construction more efficiently.(3) Based on the layer network,it is constructed a random multi-layer network model that is suitable for linear network coding transmission and the performance simulations are carried out by Visual C++.The results show that Q-Learning linear network coding can constructs the codes much faster than simple random methods... |