CN103957085B - Media access control method for wireless mesh network based on network coding - Google Patents

Abstract

The invention discloses a media access control method for a wireless mesh network based on network coding. The media access control method comprises the steps of a data package receiving stage, a data package transmitting stage and an ACK monitoring stage. According to the protocol, a network coding mechanism is introduced into a media access control layer, and throughput is increased compared with the traditional media access control method based on conflict avoidance. The media access control method is high in decoding success rate, less in control expenditure, low in calculation consumption of hardware, strong in portability, good in compatibility and capable of relieving network congestion partially.

Description

A Network Coding-Based Media Access Control Method for Wireless Mesh Networks

technical field

The invention relates to the technical field of wireless networks, in particular to a network coding-based wireless network medium access control method, which is suitable for wireless mesh networks whose service purpose is to provide high throughput.

Background technique

How to improve data throughput is a key research topic of wireless mesh network. In the case of limited channel media resources, providing high-bandwidth, high-throughput, and high-quality data transmission services for access devices places higher requirements on the physical layer and media access control layer of wireless mesh networks. At the same time, the networking equipment of the wireless mesh network is generally cheap, and the data caching capability and computing processing capability are limited. Under such harsh hardware conditions, it is also a key issue to ensure the efficient and stable operation of the medium access control method. Since network coding has a significant effect on improving throughput, and additional hardware resources are required for network coding, a medium access control method for wireless mesh networks based on network coding with stability and scalability is a breakthrough.

Wireless mesh networks have been widely used in recent years. It mainly provides wireless network access solutions for densely populated places such as enterprises, schools, shopping malls, and stations. Its main features are convenient deployment, low cost, automatic networking, and high scalability. When wired network deployment is difficult and scalability cannot be guaranteed, it is the simplest and most effective way to replace the underlying infrastructure with a wireless mesh network. The key to providing Internet access services using a wireless mesh network is how to provide a high-throughput and stable data transmission mechanism for many wireless terminals. In the prior art, in order to improve the data throughput of the wireless mesh network, there are mainly the following researches:

The first category: higher bandwidth at the physical layer; the disadvantage of this method is that it uses hardware technology to improve single-frequency bandwidth or multi-frequency multiplexing, which is technically difficult, increases the price of wireless access equipment, and has high requirements for terminal equipment. A palliative solution. At the same time, the price of high-frequency communication and multi-frequency communication in exchange for throughput is the electromagnetic pollution of the wireless signal space.

The second type: opportunistic routing at the network layer; the disadvantage of this method is that the additional communication overhead is large, the control is complicated, and the implementation is difficult. Routing protocols generally work at the network layer, and the control information used by opportunistic routing for normal operation requires additional transmission to complete, which results in additional overhead for the data link layer. Moreover, opportunistic routing is technically difficult to implement, and the consumption of hardware resources has not been studied in detail.

The third category: more stringent media access control methods; the disadvantage of this type of method is that it requires a certain degree of tacit understanding between the nodes of the wireless mesh network, that is to say, each other knows what the other is about to do, and the clock synchronization between nodes requires extremely high. Moreover, once a device in the network has a problem, other devices working with it will also face the same problem. There are problems with network universality, robustness, and scalability.

The fourth category: the network layer adds a network coding mechanism; the disadvantage of this type of method is that the network coding is independent of the media access control layer and the TCP layer, and an additional wireless broadcast protocol is required to provide reliable broadcast services. At the same time, this type of method will create a larger range of virtual bottleneck nodes in the network, resulting in large-scale network congestion. Moreover, the congestion control theory at the network layer will cause such methods to fail to work properly. In addition, this type of method seldom considers the consumption of hardware resources, so it is technically difficult to implement and has poor equipment adaptability.

Independent of the network coding mechanism of the media access control layer and the TCP layer, each node has the opportunity to listen to the data packets broadcast by neighboring nodes, and put them in its own cache for use in decoding. The defects brought in this way mainly include: 1) unrestrained monitoring of neighbor data packets is catastrophic to local cache consumption. In addition to the link load, its cache consumption is also affected by the number of neighbors. And it is unknown whether these monitored data packets will be used. The opportunity cost of throughput improvement is too high, the application uncertainty is high, and the controllability is poor. 2) The node predicts whether the neighbor has cached a certain data packet, and the probability estimation is mainly based on the ETX index of the link transmission success rate. According to the ROC verification theory, the estimation models used have a fixed error rate. When the number of nodes is high, there are many terminal devices, the communication interference is large, and the wireless link quality is poor, this fixed error rate will cause a considerable part of the coded packets to not be decoded correctly, thus wasting more hardware resources and channel resources. 3) This mechanism usually pursues encoding as many data packets as possible at a time in order to reduce the number of broadcasts of a single node. But this approach is not suitable in practice. Firstly, not all target nodes may be able to decode correctly, and secondly, when there are many target nodes, a reliable broadcast protocol will often limit node communication in a wider range, resulting in low channel utilization in the entire area.

Contents of the invention

The medium access control method of wireless mesh network based on network coding is significantly different from the usual medium access control method, aiming at the insufficiency of the existing medium access control method of wireless mesh network in terms of throughput improvement ability and many defects of the introduction of coding mechanism in the network layer , the present invention proposes a medium access control method with high throughput, low consumption, and universal robustness.

In order to realize above-mentioned task, the technical scheme that the present invention adopts is:

A network coding-based medium access control method for a wireless mesh network. In the wireless mesh network, when a node receives a data packet from a neighbor node, it judges whether the data packet is a normal data packet or a coded data packet :

If it is an encoded data packet, the encoded data packet is decoded, and the required ordinary data packet is obtained after the decoding is successful, and the encoded data packet is discarded if the decoding is unsuccessful;

If it is a common data packet, it is judged whether the common data packet reaches the generation condition of the encoded data packet, if it is achieved, it is encoded, and the encoded data packet obtained after encoding is broadcast to the neighbor node;

The generation conditions and encoding process of the encoded data packet are as follows:

In this wireless mesh network, node A has at least two neighbor nodes, if the forwarding queue of node A is empty at the initial moment, then from the initial moment: after node A receives a normal data packet p2 from the neighbor node, Look for an ordinary data packet p1 in the forwarding queue, if there is p1 that satisfies formula 1 and formula 2 at the same time, encode p1 and p2 into encoded data packet p;

(Formula 1)

In Formula 1, s1 and s2 are the source node MAC addresses of data packets p1 and p2 respectively, d1 and d2 are the MAC addresses of the next-hop nodes of p1 and p2 respectively, and t1 and t2 are the time for p1 and p2 to reach node A respectively , the value of θ is 0.3 seconds;

(Formula 2)

In Formula 2, r and c are respectively the cumulative number of normal data packets received by node A from the initial moment and the cumulative number of normal data packets satisfying formula 1 in the forwarding queue of node A, Among them, r accumulates 1 each time, c accumulates 2 each time, and the value of ξ is 0.08.

Further, ordinary data packets p1 and p2 are coded into coded data packet p through OR operation, specifically:

Do the OR operation between the MAC address of the destination node of the ordinary data packet p2 and the MAC address of the destination host of p1, and fill the result into the MAC field of the destination host of p1; do the OR operation with the MAC address of the source host MAC address of the ordinary data packet p2 and the MAC address of the source host of p1 Operation, the result is filled in the MAC field of the source node of p1; the data part of p2 and the data part of p1 are ORed, and the result is filled in the data part field of p1; if the length of the data part of p1 and p2 is inconsistent, then The tail of the shorter side is filled with 0 bits, and the new data packet generated is the encoded data packet p.

Further, in the wireless mesh network, each node is provided with a resource queue, and when a node successfully sends an ordinary data packet to another node, the ordinary data packet is moved from the forwarding queue to the resource queue, And start a countdown timer for the ordinary data packet, the initial value of the countdown timer is θ, after the countdown timer returns to zero, the ordinary data packet is removed from the resource queue.

Further, when a node receives an encoded data packet, it decodes it: the node looks for the ordinary data packet corresponding to the received encoded data packet from its own resource queue, and compares the found ordinary data packet with the received Encode the data packet to perform an OR operation to obtain the required ordinary data packet.

Further, after node A generates encoded data packet p, the steps of broadcasting p are as follows:

Step S1, create a broadcast RTS frame, write the MAC address of the next-hop node of the ordinary data packets p1 and p2 into the RTS frame, and determine the channel occupation time Tb, add Tb to the communication time field of the broadcast RTS frame, and calculate Tb Methods as below:

Tb＝5×SIFS+(82+L)×t ₀ (Formula 3)

In Formula 3, SIFS is the minimum waiting time constant specified in the 802.11 protocol, L is the length of the encoded data packet p, and _t0 is the time required to transmit a byte, which is determined by the link bandwidth;

Step S2, node A broadcasts an RTS frame to apply for a channel;

Step S3, after the neighbor node receives the RTS frame, determine the waiting time Tcts for replying the RTS frame, and reply the CTS frame to node A after the waiting time; the calculation method of Tcts is as follows:

(Formula 4)

In formula 4, iMAC is the MAC address of the neighbor node, and d1 and d2 are the MAC addresses of the next-hop nodes of data packets p1 and p2 respectively;

Step S4, node A sends the broadcast RTS frame, waits for 2×SIFS+28×t ₀ time, checks the received CTS frame, if it receives the CTS frame replied by the next hop node of p1 and p2 respectively, then node A A waits for the SIFS time and starts to send the encoded data packet p, otherwise node A enters backoff waiting.

Advantage of the present invention mainly contains the following points:

1. Increased network throughput;

Since the network coding mechanism is introduced into the data link layer, the present invention essentially improves the network throughput. Under the condition of constant hardware conditions, the use of the present invention can increase the throughput of network data by 7%-40%.

2. Strong portability and good compatibility;

Due to the good backward compatibility with CSMA/CA, the equipment supporting the present invention can communicate and form a network normally with the equipment not supporting the present invention. At the same time, the invention increases and strengthens the medium access control sublayer, does not affect other layers of the network, and the interface is transparent.

3. High decoding success rate, less additional communication;

Different from other network coding mechanisms, the present invention uses ACK frames and data packet headers to transmit necessary coding information, and nodes do not use ETX routing rules for prediction, so there is no undecodable phenomenon caused by prediction errors. Moreover, the present invention does not rely on control information of other protocols (such as opportunistic routing), and does not need to transmit independent control information between nodes.

4. Less hardware consumption;

Compared with other network coding mechanisms, this method pays more attention to the occupancy of cache resources and computing resources. Nodes no longer have the opportunity to listen to the communication between neighbors, so the cache consumption of storing decoding resources is minimized.

5. Partially solved the problem of network congestion;

In a network without congestion control, when the network load is heavy, congestion will inevitably occur at the bottleneck node of the main road and around the bottleneck node. This method just takes advantage of the large amount of information transmitted by the bottleneck node, and encodes possible data packets together. So that these data packets can quickly pass through the bottleneck node, reduce congestion, and slow down the generation of congestion.

Description of drawings

Figure 1 is a schematic diagram of the coding mechanism principle and network deployment;

Fig. 2 is a process diagram of the data sending stage;

Fig. 3 is a process diagram of the data receiving stage;

Fig. 4 is a process diagram of the ACK frame monitoring stage;

Fig. 5 (a) and Fig. 5 (b) are the structure schematic diagrams of the frame in the present invention;

Fig. 6 is a schematic diagram of the encoding data packet broadcasting process;

Figure 7 is the overall flow chart of the program;

Fig. 8 is a schematic diagram of studying the impact of network load on coding opportunities and determining ξ;

Figure 9 is a diagram of the experimental results of studying the influence of θ on the number of coding opportunities;

Figure 10(a), Figure 10(b), Figure 10(c), and Figure 10(d) are diagrams of the usage of decoding resource queues under the same ω and different θ;

Figure 11(a), Figure 11(b), Figure 11(c), and Figure 11(d) are schematic diagrams of the influence of θ on the proportions of the three nodes under different ω;

Fig. 12(a) and Fig. 12(b) are the experimental result diagrams for studying the influence of network load on the performance of each method;

Fig. 13(a) and Fig. 13(b) are the experimental result diagrams for studying the influence of link transmission success rate on the performance of each method;

Fig. 14(a), Fig. 14(b), Fig. 14(c), and Fig. 14(d) are the experimental results of studying the influence of network load on the cache usage of each method;

detailed description

The applicant deploys a large-scale wireless mesh network in buildings, in order to provide Internet access services for mobile phones, laptops and other devices, a media access control method with high throughput, good compatibility, and low resource consumption is required to ensure that all access Incoming device data quality of service. Aiming at complex situations such as limited node hardware conditions, high throughput requirements, and poor wireless channel quality, a medium access control method for wireless mesh networks based on network coding is proposed.

This method uses a network coding mechanism to improve throughput, so when deploying a wireless mesh network, except for network edge nodes, other nodes must have at least two neighbor nodes, and among all neighbor nodes, there is at least one pair that cannot communicate directly with each other neighbor nodes, as shown in Figure 1. Nodes B, C, and D are within the communication range of A, but node D is not within the communication range of B. Only in this way can node A encode the round-trip data on the link B-A-D, otherwise if D is within the communication range of B Inside, B will pass the data directly to D, and A will not get an encoding opportunity.

In this method, the positions of all nodes are fixed or only slightly move. The communication radius of all nodes remains unchanged. Three queues are set and maintained in each node: the forwarding queue is used to store data packets to be sent to other nodes, which can be ordinary data packets or encoded data packets; the resource queue is used to store data packets used for decoding Ordinary data packet; restore queue, used to store the ordinary data packet used by the restore operation after the encoded data packet fails to be sent.

The core idea of the network coding mechanism used in this method is that, as shown in Figure 1, node D first sends a normal data packet p1 to B, and then forwards it through transit node A; after A receives p1, it does not send p1 immediately Forward to B, but cache p1 for a period of time; node B sends another ordinary data packet p2 to node D, which also needs to be transferred by A; after receiving p2, A encodes p2 and p1 together, and encodes the encoded data packet p1⊕ p2 is broadcasted, and after receiving the encoded packet p1⊕p2, nodes B and D both perform decoding operations to obtain the common data packets they need. The whole process consumes three communication cycles, which is one communication cycle less than the original "receiving-forwarding" process, so as to achieve the purpose of improving throughput.

The method is based on the 802.11 protocol, and the common data packet used in the scheme is the data packet used to transmit data information between two nodes in the 802.11 protocol. In this solution, on the basis of the original 802.11 protocol, three frame formats of broadcast RTS frame, encoded data packet, and encoded ACK frame are added. The formats of various frames and data packets are basically the same as those in 802.11. A data packet ID field with a length of 4B is added to the header of the data packet frame entity, and a data packet ID field and a data packet length field are added to the header of the encoded data packet frame entity. Figure 5(a) and Figure 5(b) show all the data frame (packet) formats used. Due to the large size of the picture, it is divided into two pictures, Figure 5(a) and Figure 5(b), but it shows Table 1 takes the example in Figure 1 as an example, and provides the description of the fields with different meanings from the same data packet (frame) in 802.11 in the structure of the data packet (frame) used in this solution, except These fields and other fields have the same meanings as the fields in the 802.11 protocol. Among them, "frame" is the concept of data link layer, which is the unit of data packet. In this solution, "frame" and "packet" have the same meaning, and both refer to a bit sequence including control information and data information. This solution focuses on the discussion of the internal structure of the frame or packet, so wherever a "frame" appears, it can be replaced with a "packet", such as "encoded data frame" means "encoded data packet".

Table 1

One, the detailed steps of the inventive method

In the wireless mesh network structure of the present invention, except for the edge nodes, all other nodes carry out the data packet forwarding and processing process according to this method, and the things these nodes do are the same, so in order to clearly introduce this scheme , the following takes the model shown in Figure 1 as an example, and simplifies the scheme to an information interaction unit, that is, the information interaction process between the three nodes A, B, and D to illustrate; the A node in this model can be a mesh network Any node except the edge node, and nodes B and D are neighbor nodes of A, and D needs to send ordinary data packet p1 to B, and B needs to send ordinary data packet p2 to node D.

In this scheme, a data packet receiving phase, a data packet sending node, and an ACK frame monitoring phase are generated inside different nodes, and each phase is described in detail below.

1. Data packet receiving phase

In this scheme, the data packet sent from the source node to the destination node is called a normal data packet when the source node obtains the data packet. This data packet has the same structure as the data packet in 802.11, the only difference is A data packet ID field with a length of 4B is added to the frame entity header of the data packet, and the ID field is generated by the source node. During the transmission process of ordinary data packets at various nodes in the network, if they meet the encoding conditions at a certain node, two ordinary data packets will be encoded into encoded data packets and broadcast; if they do not meet the encoding conditions, the ordinary data packets will be According to the common data packet in the 802.11 protocol, it is forwarded to complete the transfer process at the node. Therefore, there are two types of data packets received by a node from neighboring nodes: ordinary data packets and encoded data packets.

When a node A receives a data packet from a neighbor node, it judges whether the data packet is a normal data packet or an encoded data packet, as shown in Figure 3;

Step S10, determine the type of data packet

After the node physical layer receives a complete and error-free data packet, it judges whether it is an ordinary data packet or an encoded data packet according to the length of the MAC header of the data packet. The frame is divided into three parts: frame header (Mac header), frame entity (body), and FCS field. If the length of the MAC header is 30B, it indicates that a normal data packet is received, and if the length of the MAC header is 36B, it indicates that an encoded data packet is received.

For the encoded data packet, the encoded data packet is decoded, and the decoding is successful to obtain the required ordinary data packet; if the decoding is unsuccessful, the encoded data packet is discarded; after obtaining the ordinary data packet, the node treats the data packet as a newly received data packet process;

For ordinary data packets, the node first extracts the data packet ID of the header of the data packet frame entity and records it as idx, and then searches it in its own resource queue. If a decoded resource packet containing idx is found, it means that the received normal data packet was once a part of an encoded data packet, but that encoded data packet was not successfully received. After the previous hop node decodes the encoded data packet, it performs The retransmission operation is performed, so the node should remove the found decoding resource packet. If no decoding resource package containing idx can be found, it means that the received normal data package can be encoded.

The generation conditions and encoding process of the encoded data packet are as follows:

Set the forwarding queue of node A to be empty at the initial moment, that is, there is no data packet in the forwarding queue. Each node maintains two integer variables r and c in the cache, and a four-column table b. The initial moment values of r and c are both zero, and the values of the two variables have been continuously accumulated since the initial moment; b stores auxiliary information of ordinary data packets in the forwarding queue.

The initial moment here is a moment defined for illustrating the solution of the present invention, at this moment, there is no data packet in the forwarding queue of node A. This initial moment can be understood as the time point when node A is set in the sensor network, that is, the moment when the sensor node starts to work. From this moment on, the values of variables r and c maintained in node A start from 0 and accumulate.

After node A receives the ordinary data packet p2 sent by the neighbor node B, it looks for an ordinary data packet p1 from the back to the front in the table b of the forwarding queue. Encoded into encoded data packet p;

(Formula 1)

In Formula 1, s1 and s2 are the source node MAC addresses of data packets p1 and p2 respectively, d1 and d2 are the MAC addresses of the next-hop nodes of p1 and p2 respectively, and node A can be obtained by routing algorithm; t1 and t2 are respectively The time for p1 and p2 to reach node A, the value of θ is 0.3 seconds, see Experiment 2 for the determination process.

The condition given in Formula 1 is that after node A receives the normal data packet p2, there is data packet p1 in its forwarding queue that has not been forwarded, and p1 and p2 satisfy: the next hop node (destination node) of p1 is The source node of p2, the source node of p1 is the next hop node (destination node) of p2, the time difference between p2 and p1 arriving at node A is less than θ (that is, the source node of p1 still keeps a copy of p1 in the resource queue), After these conditions are met, as shown in the example in Figure 1, D needs to send a normal data packet p1 to B, and B needs to send a normal data packet p2 to node D;

(Formula 2)

In formula 2, r is the cumulative number of ordinary data packets received by node A from the initial moment, that is, as long as A receives an ordinary data packet, the value of r will be increased by 1; c refers to node A In the forwarding queue, the cumulative number of ordinary data packets satisfying formula 1 is the number of data packets that can be encoded. If an ordinary data packet meets the conditions of formula 1 and can be encoded, the value of c is accumulated by 2; ξ is a fixed constant with a value of 0.08; k is an encoding switch. When k<ξ, it means that the current node is working under low load, or the nearby network is providing services for FTP, streaming media and other applications, then the Ordinary data packets do not need to be encoded, and are forwarded according to 802.11 ordinary data packets; and when k>ξ, it indicates that the direction of data flow through the current node is sufficiently complicated, and this method should be used for encoding of ordinary data packets, and the determination process of ξ value See experiment one.

Step S11, encoding of common data packet

In this scheme, ordinary data packets p1 and p2 that meet the coding conditions are encoded into coded data packets p through OR operation, that is, p=p1⊕p2, specifically:

Node A puts the ordinary data packet p2 into the restoration queue, which is used in the restoration process when the encoded data packet p fails to be sent. Node A adds a destination node 2MAC field after the destination node MAC field in the MAC header of p1, and writes the destination node MAC address of p2 in it; performs an OR operation with the destination host MAC address of p1 and the destination host MAC address of p2, and the result is The result is filled in the destination host MAC field of p1; the OR operation is performed with the source host MAC address of p1 and the source host MAC address of p2, and the result is filled in the source host MAC field of p1; after the frame entity data packet ID field of p1, Add the packet ID of p2, the length of the frame entity data part of p1, and the length of the frame entity data part of p2; use the frame entity data part of p1 and the frame entity data part of p2 to do OR operation, and fill the result into the frame entity data part of p1 field, if the lengths of the data parts of p1 and p2 are inconsistent, fill the tail of the shorter side with 0; then the new data packet generated after encoding the two is the encoded data packet p.

Step S12, coded information transmission

After node A encodes the ordinary data packets p2 and p1 together, it makes an encoded ACK frame, the suffix ID1 field of the frame is filled with the data packet ID in the header of the data packet p1 frame, and the suffix ID2 field is filled with the data packet p2 frame The packet ID of the header. After waiting for SIFS time, Node A sends the encoded ACK frame to Node B.

Step S13, decoding the encoded data packet

After making the encoded data packet p, A broadcasts it to the neighbor nodes. When node D sends the data packet p1 to node A, node D moves p1 from the forwarding queue to the resource queue, and starts a countdown timer for the ordinary data packet p1. The initial value of the countdown timer is θ, and the countdown timer resets to zero , remove the normal packet from the resource queue. The ordinary data packet p1 moved into the resource queue is used when D receives the encoded data packet generated by p1 for decoding.

Its neighbor node D receives the p broadcasted by A, first extracts the data packet 1 ID, data packet 2 ID, data packet 1 length, and data packet 2 length from the header of the frame entity of p, which are respectively recorded as id1, id2, l1, l2. Node D looks for the normal data packet corresponding to the received coded data packet p in its own resource queue, that is, the normal data packet whose frame entity header contains id1 or id2. After finding it, record this ordinary data packet as p3. If p3 cannot be found, it means that p cannot be successfully decoded, and p is discarded.

Node D extracts the data packet ID and data packet length from p3, which are recorded as id3 and l3 respectively. If id3 is equal to id2, then copy the MAC address of the destination node 2 of the encoded data packet p to the MAC field of the destination node of p3; use the MAC address of the destination host MAC address of p3 and the MAC address of the encoded destination host of p to do OR operation, and put the result into p3 The destination host MAC field of p3; use the source host MAC address of p3 and the coded source host MAC address of p to do the OR operation, and the result is put into the source host MAC field of p3; use the data part of the p3 frame entity and the p coded frame entity If the coded data part in is ORed, if the length of the obtained result is greater than l2, it will be intercepted, only the first l2 bytes will be kept, and the intercepted result will be written into the frame entity data part of p3. When id3 is equal to id1, the method is similar; the newly produced p3 is the decoded data packet p2.

When node D decodes p, it determines the reply ACK frame waiting time Tack, and then replies a normal ACK frame to node A, indicating that it has received the encoded data packet and has successfully decoded the data packet. After node D successfully decodes p, it submits the processed p3 data packet to the network layer.

Note: The decoding process of the encoded data packet introduced in step S13 is explained with node D as the current node, and if node A receives the encoded data packet sent by other nodes, it is the same as the decoding process of this part. Taking D as the current node to explain is based on the example given in Figure 1. In fact, in the application process, the process completed by node A occurs in every node except the edge node. B-A-D is A simple model abstracted to explain this scheme, in fact, the completion process of B, D and A is the same.

2. Data packet sending phase

Step S20, determining the channel occupancy type

Node A judges the length of the MAC header of the first data packet in its forwarding queue. If it is 30B, it means that this is an ordinary data packet and operates according to the RTS-CTS process of 802.11.

If the length of the MAC header is 36B, it means that an encoded data packet will be broadcast, such as the encoded data packet p, as shown in 2.

Step S21, creating an RTS frame

Create a broadcast RTS frame, add your own MAC address to the source node MAC field of the frame, copy the destination node 1MAC and destination node 2MAC fields in the MACheader of p to the corresponding fields of the encoded RTS frame, as shown in Figure 1 Take the model as an example, that is, copy the MAC address of the next-hop node of the ordinary data packets p1 and p2 to the MAC field of the broadcast RTS frame; then determine the channel occupation time Tb, and add Tb to the communication time field of the RTS frame. The calculation method of Tb is as follows :

Tb＝5×SIFS+(82+L)×t ₀ (Formula 3)

In Formula 3, SIFS is the shortest waiting time constant specified in the 802.11 protocol, L is the length of the encoded data packet p, and t ₀ is the time required to transmit a byte, which is determined by the link bandwidth; Schematic diagram of the process, Tb is the time elapsed from node A sending a broadcast RTS frame to node B ending sending an ACK frame, including 5 SIFS times, 1 broadcast RTS frame transmission time, 2 CTS frame transmission times, and 1 encoding Data frame transmission time and 2 normal ACK frame transmission times. The length of broadcast RTS frame is 26B, and the length of CTS frame and normal ACK frame is 14B.

Step S22, node A broadcasts an RTS frame to apply for a channel;

Step S23, after the neighbor node receives the RTS frame, determine the waiting time Tcts for replying the RTS frame, and reply the CTS frame to node A after the waiting time; the calculation method of Tcts is as follows:

(Formula 4)

In Formula 4, iMAC is the neighbor node. If the neighbor node D receives the RTS frame, it is the MAC address of D, and d1 and d2 are the MAC addresses of the next-hop nodes of the data packets p1 and p2, namely B and D The MAC address of node D; after waiting for Tcts time, if node D finds that no nodes around are using the channel, it will reply a CTS frame to node A, indicating that it can communicate with node A. In the next Tb-Tcts-40×t ₀ time, no reply will be made even if the RTS frame sent by other nodes is received. If node D finds that the surrounding channels are being used, it directly discards the RTS frame sent by node A. Node B does the same thing as D.

Step S24, determine the channel application result

Node A sends the broadcast RTS frame, waits for 2×SIFS+28×t ₀ time, and checks the received CTS frame. If it receives the CTS frame replied by the next-hop nodes of p1 and p2 respectively, it means that the channel application is successful. Then node A waits for the SIFS time and starts to send the coded data packet p. If only one CTS frame is received or no CTS frame is received, the instruction channel application fails, and node A enters backoff waiting.

After node D receives the encoded data frame broadcast by A, it performs a decoding operation (see step S13 for the specific process), and determines the waiting time Tack for replying to the ACK frame. The calculation method of Tack is as follows:

Among them, iMAC is the MAC address of node D, MACheader.DtMAC1 is the MAC address of destination node 1 of the encoded data packet received by node D, that is, the MAC address of B; MACheader.DtMAC2 is the MAC address of destination node 2, that is, the MAC address of D. Node D waits for the Tack time, and if the encoded packet is decoded correctly, it will reply a normal ACK frame to node A, otherwise it will directly discard the received encoded data packet. Node B does the same thing as D.

Step S25, determine the retransmission status

Node A waits for 2×SIFS+28×t ₀ time after sending p=p1⊕p2 to check the received ACK frame. If node A receives two ACK frames from node B and node D respectively, it means that B and D have received the encoded data packet p1⊕p2 and successfully decoded it. At this time, node A removes p from the forwarding queue and removes p from its own Find p2 in the restoration queue of , and remove p2. If an ACK frame is not received, it enters backoff waiting.

If only the ACK frame replied by node D is received, but the ACK frame replied by B is not received, it means that B fails to receive the encoded data packet p correctly or cannot decode it correctly. At this time, node A needs to find the normal data packet p2 from its restore queue, directly replace p1⊕p2 in the forwarding queue with p2, and then back off and wait.

If only the ACK frame replied by node B is received, but not the ACK frame replied by D, it means that B failed to receive the encoded data packet p1⊕p2 correctly or could not decode it correctly. Node A finds the normal data packet p2 from its restore queue. Node A needs to remove the MAC field (address) of the destination node 2 in the MAC header of the encoded data packet p1⊕p2; use the MAC address of the encoded destination host and the MAC address of the destination host of p2 to perform an OR operation, and replace the encoding of p with the obtained result The MAC address of the destination host; use the MAC address of the encoded source host and the MAC address of the source host of p2 to do an OR operation, and replace the encoded source host MAC address of p with the obtained result; remove the packet 2ID and the packet in the header of the encoded packet frame entity 2 length; use the coded data part in the frame entity of p and the data part in the frame entity of p2 to perform an OR operation, and replace the coded data part of p with the obtained result; truncate the coded frame according to the length of the remaining data packet 1 in the coded frame entity header Entity, discard the excess; refresh the FCS field of p; finally remove p2 from the restore queue. In this way, the encoded data packet p is restored to data packet p1, and then backoff waits.

3. ACK frame monitoring stage

The ACK frame monitoring phase in the embodiment of the present invention includes updating resource queues, maintaining resource timers, and the like. The main function is to manage the decoding resource queue. At this stage, the node not only listens to the ACK frame that will be sent to itself, but also listens to the ACK frame passed between other nodes during the backoff waiting process.

Fig. 4 is a process diagram of the ACK frame monitoring stage, and in combination with Fig. 1, Fig. 5 and Fig. 6, the specific steps of the process diagram of the ACK frame monitoring stage in this embodiment are as follows:

Step S30, determining the state of the ACK frame. After the node listens to the ACK frame, it checks whether the destination node MAC field of the ACK frame is the same as its own MAC address. If they are the same, it means that the ACK frame is an acknowledgment of the data packet you just sent, and you need to update the resource queue and create a resource timer; if they are not the same, it means that the ACK frame is obtained by idle listening, check whether the ACK frame is encoded ACK frame, that is, whether it contains the suffix ID field, if it does not contain the suffix ID field, the ACK frame is directly discarded, otherwise, the resource timer needs to be maintained.

Step S31, updating the resource queue. After node B sends the normal data packet p2 to node A, it waits for SIFS. If no ACK frame is received, it means that p2 fails to be sent and enters backoff waiting. If an ACK frame is received, regardless of whether the ACK frame is an encoded ACK frame or not, it means that A has successfully received p2. Node B moves the normal data packet p2 from the forwarding queue to the resource queue, and starts a countdown timer for the resource packet p2. The initial value of the countdown timer is θ. After the countdown timer returns to zero, the resource packet p2 is removed from the resource queue. Because according to formula 1, the countdown timer is reset to zero, and node A will not encode p2 again.

Step S32, check the encoding information. If the ACK frame received by node B is an encoded ACK frame, it means that node A not only successfully received the ordinary data packet p2, but also encoded p2 with other data packets. Node B extracts the data packet ID from the suffix ID1 of the ACK frame and records it as id1, that is, the data packet ID of the frame entity header of the ordinary data packet p1. The node B stops the countdown timer started for the resource packet p2 in step S31 to prevent p2 from being accidentally removed, causing the encoded data packet p1⊕p2 to not be decoded. Node B changes the data packet ID of the frame entity header of the normal data packet p2 to id1, so as to operate in step S13. Thereafter, there are only two situations in which the ordinary data packet p2 is removed from the resource queue: one is that the coded data packet p arrives and is decoded successfully, and the other is that the data packet whose frame entity header is id1 arrives (step S13).

Step S33, updating the countdown timer. After node D sends the ordinary data packet p1 to node A, it receives the confirmation ACK frame replied by node A, which is an ordinary ACK frame. According to step S31, p1 is moved to the resource queue of node D, and a countdown timer is running for resource Packet p1 timing. Since then, although node D does not use the channel to send data, it has to listen to the communication between other nodes within the communication range. After node A receives the ordinary data packet p2 sent by node B and meets the encoding conditions, it encodes p2 and replies an encoded ACK frame to node B. At this time, node D will listen to the coded ACK frame idle. Node D extracts the suffix ID1 of the encoded ACK frame as id1, and extracts the suffix ID2 as id2. Node D searches the resource queue for a resource packet whose frame entity header is id1. If it finds it, it means that the coded ACK frame detected by idle monitoring is related to itself, and changes the found resource packet frame entity header to id2, and stops this resource packet. countdown timer; if it cannot be found, it means that the coded ACK frame has nothing to do with itself, and it will be discarded directly.

Two, the determination of each relevant parameter in the inventive method:

Experiment 1: Study the influence of network load on encoding opportunities and the determination of ξ;

Step 1, simulation experiment scene initialization:

The applicant simulated a wireless mesh network with a length of 1400m and a width of 600m, in which 86 nodes are arranged. In this experiment, the communication radius of each node is 175m, and the link transmission success rate between nodes is measured by the physical environment. The simulation time of each experiment is 5s, each node generates ω data packets per second, the destination address is randomly selected, the length of each data packet is 1000B, and the link bandwidth is 54Mbps. The network layer uses the minimum number of hops for routing. ω controls the network load.

Step 2, take ω=50, 100, 150...300, and for each value of ω, take θ=0.25s, 0.50s, 0.75s, 1.00s, therefore, a total of 6×4 groups of evaluations are done. In order to ensure the authenticity of the experimental results, 1000 experiments were carried out for each group of evaluations, and the expected value of the results was taken as the final result of each group.

Step 3, analyze and process experimental data:

Fig. 8 shows that under different ω values, in step S23 (data flow complexity) changes. Different lines in the figure represent different values of θ. It can be seen from the experimental results: (1) For different θ, the data flow complexity has the same trend. (2) When ω<50, the complexity of data flow increases rapidly; as ω increases, the complexity of data flow decreases gradually, and then tends to be stable. (3) The data flow complexity curves of different θ are very close. These three observations show that when the network load is small, nodes can easily obtain channel resources, and there are few data packets in the forwarding queue, and there are almost no data packets that can be encoded together; as the network load increases, the forwarding queue Start to pile up data packets, there are many cases where two data packets can be encoded together, that is to say, the number of encoding opportunities is increasing; however, the total number of data packets that can be encoded together accounted for the proportion of received data packets is not always increasing Yes, when the network load reaches a very high value, the data flow complexity tends to a constant, which is related to the network bandwidth and deployment. That is to say, when the network load is high, the factor affecting the complexity of the data flow is mainly the complexity of the data flow in the network. Therefore, we observed that when When <0.08, there are only two possible situations: one is that the node and its neighbors are working under low load conditions, and the forwarding queue is basically empty; the other is that the node and its neighbors are providing services for some special applications such as FTP and streaming media , the complexity of the data flow in the network is relatively low, basically one-way. In order to reduce the algorithm complexity, we determined ξ to be 0.08.

Experiment 2: Study the effect of θ on the number of encoding opportunities;

Step 1, simulation experiment scene initialization:

The applicant simulated a wireless mesh network with a length of 1400m and a width of 600m, in which 86 nodes are arranged. In this experiment, the communication radius of each node is 175m, and the link transmission success rate between nodes is measured by the physical environment. The simulation time of each experiment is 5s, each node generates ω data packets per second, the destination address is randomly selected, the length of each data packet is 1000B, and the link bandwidth is 54Mbps. The network layer uses the minimum number of hops for routing. ω controls the network load.

Step 2, take θ=0.25s, 0.50s, 0.75s, 1.00s, take ω=10, 20, 30, 40, 50, 100, 150, 200, 250, 300, and conduct a total of 4×10 groups of evaluations. In order to ensure the authenticity of the experimental results, 1000 experiments were carried out for each group of evaluations, and the expected value of the results was taken as the final result of each group.

Step 3, analyze and process experimental data:

Figure 9 shows the change of coding opportunities under different θ. Different lines in the figure represent different values of ω. From the experimental results, it can be seen that: (1) when ω<50, the coding chance hardly changes with the increase of θ; (2) when ω>50, the coding chance increases with the increase of θ. These two observations show that when the network load is small, even if θ is large, there will be no more coding opportunities, and the main node can easily apply for the channel to send the data to be forwarded; when the network load is large, as θ The coding opportunities are gradually increasing, and the growth trend under different loads is basically the same. This growth acceleration is determined by the complexity of the data flow. In addition, the vertical distances of the lower 6 lines in Figure 9 are relatively large, much larger than the vertical distances of the upper 4 lines, indicating that network complexity has a greater impact on encoding opportunities. The final conclusion from this experiment is that θ has a far less impact on encoding opportunities than ω, because θ cannot be set very large in order to obtain more encoding opportunities.

Experiment 3: Study the influence of θ on node cache occupancy, and the determination of θ;

Step 1, simulation experiment scene initialization:

The applicant simulated a wireless mesh network with a length of 1400m and a width of 600m, in which 86 nodes are arranged. In this experiment, the communication radius of each node is 175m, and the link transmission success rate between nodes is measured by the physical environment. The simulation time of each experiment is 5s, each node generates ω data packets per second, the destination address is randomly selected, the length of each data packet is 1000B, and the link bandwidth is 54Mbps. The network layer uses the minimum number of hops for routing. ω controls the network load. The cache allocation of each node is as follows: the length of the forwarding queue is not limited, the length of the decoding resource queue is 400, if the decoding resource queue is used up, there is another emergency queue with a length of 500 for storing decoding resources, recovery queue occupancy and decoding resource queue usage There is correlation, so no key discussion will be made.

Step 2: Take θ from 0.1s to 1.0s, with a step of 0.1s, and ω from 50 to 300, with a step of 50. A total of 10×6 groups of evaluations are performed. In order to ensure the authenticity of the experimental results, 1000 experiments were carried out for each group of evaluations, and the expected value of the results was taken as the final result of each group.

Step 3, analyze and process experimental data:

Figure 10(a-d) shows the resource queue usage under the same ω and different θ. Points of different shapes in the figure represent nodes with different consumption levels: square nodes consume serious resources, not only exhausting the resource queue, but also using emergency queues; circular nodes use normal resources; diamond-shaped nodes use less resources and have low utilization. From the experimental results, it can be seen that: (1) when θ is small, the resource queues of most nodes are not fully utilized; (2) when θ is large, although the resource queues of most nodes are fully utilized, the nodes that exhaust the resource queues There are also many.

Figure 11(a–d) shows the effect of θ on the proportion of three kinds of nodes under different ω. From the experimental results, it can be seen that: (1) when the network load is small, as θ increases, the distribution ratio of the three nodes remains basically unchanged; (2) when the network load is in the middle, compared with the case of light load, the cache The nodes with low utilization rate are greatly reduced, and all of them are converted into normal nodes, but at the same time, under this moderate load, as θ increases, some normal nodes will convert into red sick nodes. When θ>0.3s, This change is particularly obvious; (3) when the network load is high, there are fewer nodes with low cache utilization, but the number of such nodes also begins to be affected by θ, and the number of pathological nodes is more than moderate load. This is due to the fact that under high load conditions, the amount of data will be dispersed in a wider range, rather than quickly collected to some nodes.

These observations indicate that θ has little effect on the cache usage when the network load is low, which is the network deployment and link quality which are the main factors affecting the cache utilization. When the network load is high, The choice of θ will bring about two main effects, one is to affect the number of low-utilization nodes, and the other is to affect the number of sick nodes. From this experiment, the final result is that if you want to increase the cache utilization and control the sick nodes within the tolerance range, θ is set to 0.3s.

Three, the comparative experiment of the inventive method and other methods

Next, we verify the advantages of the method of the present invention over other protocols through a set of experiments. The experiment mainly compares the performance of the following three algorithms:

(1) FSNC: the method proposed by the present invention.

(2) CSMA/CA: This protocol is the most widely used media access control method in 802.11. It uses the RTS-CTS mechanism for collision avoidance. There is no reliable broadcast module and no encoding mechanism.

(3) XORs: This method is the first classic case of combining network coding with practical applications, and it is also the most researched and discussed method. Different from the method proposed by the present invention, XORs puts the network coding mechanism between the media access control layer and the TCP layer, and makes no modification to both layers.

In these experiments, we used the concept of information throughput as a measurement indicator, which means: the sum of the information content of the data packets sent by all nodes in the entire network per unit time, where the information content refers to the information content that is involved in encoding. The number of data packets for OR operation, and the information content of unencoded data packets is always 1. The experiment mainly proves the advantages of the present invention from the following aspects:

①The impact of network load on protocol performance, ②The impact of link transmission success rate on protocol performance, ③The impact of network load on cache utilization.

(1) The impact of network load on protocol performance:

Simulation network initialization:

The applicant simulated a wireless mesh network with a length of 1400m and a width of 600m, in which 86 nodes are arranged. In this experiment, the communication radius of each node is 175m, and the link transmission success rate between nodes is measured by the physical environment. The simulation time of each experiment is 5s, each node generates ω data packets per second, the destination address is randomly selected, the length of each data packet is 1000B, and the link bandwidth is 54Mbps. The network layer uses the minimum number of hops for routing. ω controls the network load. According to the previous test results, we set the parameters θ to 0.3s and ξ to 0.08.

Simulation experiment process:

In this experiment, ω=10, 20, 30, 40, 50, 100, 150, 200, 250, 300 were taken, and a total of 10 groups of evaluations were performed. In order to ensure the authenticity of the experimental results, 1000 tests were carried out for each method and each group of evaluations, and the expected value of the results was taken as the final result of each group.

Experimental results:

Figure 12(a, b) shows the relationship between the network-wide many-to-many throughput of each protocol and the network load ω. It can be seen from the figure that as the network load increases, the throughput of CSMA/CA without a coding mechanism gradually increases and then stabilizes. This is because when the network load is high, the channel bandwidth is exhausted, and the throughput of the entire network will not Infinite increase; compared with CSMA/CA, FSNC has a more obvious throughput improvement, which is basically maintained at more than 7%, and can reach 40% at the most; when the network load is between 50 and 100, the throughput of XORs is higher than that of FSNC , this is because the main difference between XORs and FSNC is the use of opportunistic interception, neighbor state prediction based on ETX routing rules, and encoding of multiple packets together. However, under low and high load conditions, the throughput of XORs is not as good as FSNC. This is because, when the load is low, XORs cannot find opportunities to encode multiple packets together and give up encoding. When the load is high , some nodes encode multiple data packets together at one time and broadcast to more objects, but because XORs also explicitly uses a reliable broadcast protocol, it makes it very difficult to apply for a channel when broadcasting encoded packets, often due to a certain node channel The entire broadcast process is canceled due to busyness; in addition, when a data packet with a large amount of information is broadcast, all neighbors of the receiving node cannot use the channel, which leads to low utilization of channel resources in a wider range, thus affecting the throughput of the entire network quantity. This experiment shows that FSNC has higher throughput than CSMA/CA, and has better adaptability than XORs, and can adapt to various network loads.

(2) The impact of link transmission success rate on protocol performance:

Simulation network initialization:

The applicant simulated a wireless mesh network with a length of 1400m and a width of 600m, in which 86 nodes are arranged. In this experiment, the communication radius of each node is 175m, and the simulation time of each experiment is 5s. Each node generates ω data packets per second, and the destination address is randomly selected. The length of each data packet is 1000B. The bandwidth is 54Mbps, and the network layer uses the minimum number of hops to route, and controls the network load through ω. According to the previous test results, we set the parameters θ to 0.3s and ξ to 0.08.

Simulation experiment process:

In this experiment, take ω=100,300, use random generation, and replace the actual measured link transmission success rate with the link transmission success rate that satisfies the Gaussian distribution. The expected value of the Gaussian distribution ranges from 1 to 0.8, and the jump step is 0.02. A total of 2 sets of evaluations were carried out. In order to ensure the authenticity of the experimental results, 1000 tests were carried out for each set of evaluations for each protocol, and the expected value of the results was taken as the final result of each set.

Experimental results:

Figure 13(a,b) shows the relationship between the network-wide many-to-many throughput of each protocol and the link transmission success rate under moderate load and high load. It can be seen from the figure that in the case of moderate load, as the channel environment deteriorates, both CSMA/CA and FSNC need to pay extra information throughput to overcome the error packet loss phenomenon caused by the channel environment, while XORs in When the channel environment is good, there is a higher throughput, because in this case, the results obtained by the XORs neighbor prediction model are correct in a high proportion, but as the channel environment becomes worse, the accuracy rate of the XORs prediction model It began to decrease, many encoded packets could not be successfully decoded, and the throughput began to approach FSNC. In the case of high load, the throughput of XORs and CSMA/CA is very close. This is because when XORs broadcasts high-informative encoded packets, all the neighbors of the receiving node are silenced, resulting in low regional channel utilization, which affects the The throughput of the entire network; moreover, the throughput of these two protocols hardly changes with the change of link quality, because when the load is high, the channel can no longer provide additional bandwidth for data packet retransmission. At this point, FSNC still maintains a 7% increase in throughput. This experiment shows that the performance of FSNC is less affected by the success rate of link transmission and more robust.

(3) The impact of network load on cache utilization:

Simulation network initialization:

The applicant simulated a wireless mesh network with a length of 1400m and a width of 600m, in which 86 nodes are arranged. In this experiment, the communication radius of each node is 175m, and the link transmission success rate between nodes is measured by the physical environment. The simulation time of each experiment is 5s, each node generates ω data packets per second, the destination address is randomly selected, the length of each data packet is 1000B, and the link bandwidth is 54Mbps. The network layer uses the minimum number of hops for routing. ω controls the network load. According to the previous test results, we set the parameters θ to 0.3s and ξ to 0.08.

Simulation experiment process:

In this experiment, ω=60, 100, 200, 300 were used, and a total of 4 groups of evaluations were carried out. In order to ensure the authenticity of the experimental results, 1000 tests were carried out for each group of evaluations for each protocol, and the expected value of the results was taken as the final result of each group.

Experimental results:

Figure 14(a-d) shows the relationship between the usage of XORs and FSNC decoding resource cache queues and network load. It can be seen from the figure that under various network loads, the cache usage of XORs nodes is higher than that of FSNC, and even several times higher in some cases. This is because XORs nodes not only cache the data forwarded by themselves, but also cache opportunities. The monitored data of neighbor nodes does not have a good cache management mechanism. In addition, XORs points are more widely distributed than FSNC points, and FSNC points are more concentrated, which means that the cache usage of XORs is quite different, the method is less stable, and it is more difficult to control the allocation of caches. In contrast, FSNC is more stable , is also easier to control.

Claims (4)

1. A method for controlling medium access in a wireless mesh network based on network coding, characterized in that, in the wireless mesh network, after a node receives a data packet sent by a neighbor node, it is judged that the data packet is normal Packets or encoded packets: If it is an encoded data packet, the encoded data packet is decoded, and the required ordinary data packet is obtained after the decoding is successful, and the encoded data packet is discarded if the decoding is unsuccessful; If it is an ordinary data packet, it is judged whether the ordinary data packet reaches the generation condition of the encoded data packet, if it is achieved, it is encoded, and the encoded data packet obtained after encoding is broadcast to the neighbor node; if the generation condition cannot be met, The ordinary data packet is then forwarded according to the 802.11 protocol; The generation conditions and encoding process of the encoded data packet are as follows: In this wireless mesh network, node A has at least two neighbor nodes, if the forwarding queue of node A is empty at the initial moment, then from the initial moment: after node A receives a normal data packet p2 from the neighbor node, Look for an ordinary data packet p1 in the forwarding queue, if there is p1 that satisfies formula 1 and formula 2 at the same time, encode p1 and p2 into encoded data packet p; In Formula 1, s1 and s2 are the source node MAC addresses of data packets p1 and p2 respectively, d1 and d2 are the MAC addresses of the next-hop nodes of p1 and p2 respectively, and t1 and t2 are the time for p1 and p2 to reach node A respectively , the value of θ is 0.3 seconds; In Formula 2, r and c are respectively the cumulative number of normal data packets received by node A from the initial moment and the cumulative number of normal data packets satisfying formula 1 in the forwarding queue of node A, Among them, r accumulates 1 each time, c accumulates 2 each time, and the value of ξ is 0.08; The ordinary data packets p1 and p2 are coded into coded data packets p through an OR operation, specifically: Do an OR operation with the MAC address of the destination host of the ordinary data packet p2 and the MAC address of the destination host of p1, and fill the result into the MAC field of the destination host of p1; perform an OR operation with the MAC address of the source host of the ordinary data packet p2 and the MAC address of the source host of p1 , the result is filled in the source host MAC field of p1; the data part of p2 is ORed with the data part of p1, and the result is filled in the data part field of p1; if the length of the data part of p1 and p2 is inconsistent, the length The shorter side is filled with zeros at the end, and the new data packet generated is the coded data packet p. 2. The medium access control method of wireless mesh network based on network coding as claimed in claim 1, characterized in that, in the wireless mesh network, each node is provided with a resource queue, and when a node sends an ordinary data After the packet is successfully sent to another node, move the ordinary data packet from the forwarding queue to the resource queue, and start a countdown timer for the ordinary data packet. The initial value of the countdown timer is θ. After the countdown timer is reset to zero, the ordinary The packet is removed from the resource queue. 3. The medium access control method of wireless mesh network based on network coding as claimed in claim 2, characterized in that, after a node receives a coded data packet, it is decoded: the node searches for Ordinary data packets corresponding to the received encoded data packets are ORed by the found ordinary data packets and the received encoded data packets to obtain the required ordinary data packets. 4. the wireless mesh network medium access control method based on network coding as claimed in claim 1, is characterized in that, after described node A generates coded data packet p, the step of broadcasting p is as follows: Step S1, create a broadcast RTS frame, write the MAC address of the next-hop node of the ordinary data packets p1 and p2 into the RTS frame, and determine the channel occupation time Tb, add Tb to the communication time field of the broadcast RTS frame, and calculate Tb Methods as below: Tb＝5×SIFS+(82+L)×t ₀ (Formula 3) In Formula 3, SIFS is the minimum waiting time constant specified in the 802.11 protocol, L is the length of the encoded data packet p, and _t0 is the time required to transmit a byte, which is determined by the link bandwidth; Step S2, node A broadcasts an RTS frame to apply for a channel; Step S3, after the neighbor node receives the RTS frame, determine the waiting time Tcts for replying the RTS frame, and reply the CTS frame to node A after the waiting time; the calculation method of Tcts is as follows: In formula 4, iMAC is the MAC address of the neighbor node, and d1 and d2 are the MAC addresses of the next-hop nodes of data packets p1 and p2 respectively; Step S4, node A sends the broadcast RTS frame, waits for 2×SIFS+28×t ₀ time, checks the received CTS frame, if it receives the CTS frame replied by the next hop node of p1 and p2 respectively, then node A A waits for the SIFS time and starts to send the encoded data packet p, otherwise node A enters backoff waiting.