Network code have, employ well in the intersection of broadband and wireless self-organization network as one new technology, through the network code, the intermediate node can encode and send received information out, have improved network throughput and robustness. Do not do very great modification to soft hardware device and corresponding agreement of the existing network, can choose to realize the network code in the high level. Wireless sensor network, wireless grid network (Mesh) Can use the encoding technique of the network to improve the transmission performance of the periodic line of multihop notably while waiting for the wireless self-organization network. The coded research focus of the network concentrates on the code node of the network and chooses the scheme at present, algorithmic design of network code, code complexity analysis of network, network coded perfomance analysis, network code and safety analysis of the system, network code are in such aspects as application in the wireless distributed network.
The design object of the wireless multihop communication system of broadband is on the premise of fully utilizing the definite wireless network resource, enable every receiving node to be fast to receive the intact message. How to raise the characteristic of the self-organisation wireless network of multihop, it has been the industry research and focal point [1 ] of the concern all the time.
A network encoding technique principle
Network code (Network coding) Come, say it is node in the network that go on, transmit multicast (Multicast) that go out encode, message that receive from broad sense Technology. Multicast (call multicasting too) It is an important communication mode in the network. At at the same time to several other nodal transmission data one or several node, will often rely on other nodal transmission.
In the traditional network, can only duplicate, amplify and transmit the signal received as the node relaying, sometimes to network resource this is a kind of waste. The encoding technique of the network has broken this kind of restriction, the message that it allows to relay the pair of nodes and receive is encoded, and send a plurality of data packets received out according to recombination of a certain particular algorithm.
Fig. 1 shows as the topological embodiment of a wireless communication domain 3 node: Node A, node B transmit information a, b each other. The arrowhead in Fig. 1 represents the directed periodic line, suppose the capacity of each periodic line is " 1 " . Fig. 1(a) Adopt the traditional communication mode, A is to S transmitted information a at first, then B is to S transmitted information b, S then differentiates radio of information a and b for the node A and node B sequentially. Can obtain information a the transmission node B of 4 periodic lines so, and the node A can obtain information b. But when information a and b plans to transmit through the node S, if employ the encoding technique of the network, make the operation of modulo-2 sum of a and b, transmit out directly, locate in node B, can resume a according to the message received; Same, at node A too can resume, happen information come by b, thus can decipher receive the intersection of information and b. After adopting the encoding technique of the network (refer to fig. 1(b) ),Only need to use 3 periodic lines to realize all communication of the traditional way is required.
Can be found out from the embodiment, the encoding technique of the network can be by improving the data rate of the multicast prominently. The information transmitted in a network, can be by so-called " flows " of the image . The peak rate of flow of the network is that the maximum from source point to sink transmits the data rate. And peak rate of flow that broadcast mean at source point at the same time to send same data by all sink, each maximum message transmission rate which can receive in sink. Theoretically, maximal flow depends on the topological structure of the network, namely every nodal connection relation and bandwidth. Famous maximal flow is minimum can get the maximal flow [2 ] to the sink of a certain source point in the given network to cut the theorem in the graph theory.
Maximal flow is minimum to cut the theorem: Any tape sends some and all there is maximal flow and minimum in the network of the sink to cut, and the rhe of maximal flow equals the minimum capacity that cuts.
Can also be found out from Fig. 1, an admissible network coding scheme, must make the sink resume raw information from the data that are received, that is to say, according to the data received, and known coding scheme, can solve out the original data onlily. If regard the whole network as a system, the data that the source point sends can be regarded as systematic input, the data received in sink are as exporting, each operation with nodal code of the middle is regarded as the system function, require that full rank from the information transmission matrix of the source point between each sink, can meet the request of broadcasting.
The data may pass numerous codes while passing the go-between. A node if oneself is not a source point, then the message that it sends out can only come from the message that receive. So encode in any case, the information volume that it sends out must be no more than the information volume that is received. Say, the data pass a node each time in the course of transmitting in terms of information theory, their information entropy is not increased. So, do not reduce for the information entropy that guarantees to be passed to the sink finally and require every intermediate node is while encoding, the transmission matrix of the system does not lower the order, namely can't harm codes. And this node, in order to judge whether the present coding scheme will cause systematic transmission matrix to lower the order, also know various algorithmic improvements of other nodal places.
Because each node is applied is the stationary code vector for the deterministic coding scheme, so only need to include the information vector in the data transmitted in the network, save the bandwidth, and the required set of symbol is smaller; But the deterministic network encodes the situation needing to find out about the whole network, complexity is higher, it is difficult to realize distributedly. Once the network topology changes, must revise the whole coding scheme, the robustness is worse.
(2)Random coding scheme
Because the above coded shortcoming of the determinacy network, Ho and Medard,et al. [7 ] have put forward the coded concept at random, random coding is a node that lets in the network in totally independent distributed way, the random selection encodes the coefficient, encode input information and and send the random vector of this group to the sink as a part of header, in order to decode. Has already proved, it is at the time of the infinity in the set of symbol, adopt random coding, the probability that systematic transmission matrix is full rank is " 1 " .
