CN110086518B - Elastic beamforming method based on multi-arm gambling machine in wireless sensor network - Google Patents

Abstract

The invention relates to a multi-arm gambling machine-based elastic beam forming method, which comprises the following steps: m sensor nodes are arranged in the wireless sensor network; when there is fault sensor node interference, the set containing different sensors is defined as an arm, and then K groups of arms are shared, wherein K is 2^m-2; initializing the selected times of each arm, the estimated value of each arm, and the probability of each arm being selected at the initial moment_i1/K; randomly selecting one arm according to the selected probability, carrying out phase updating on all sensor nodes in the selected arm once according to a random grouping distributed beam forming method, increasing the selected times corresponding to the selected arm by one, updating a reward value corresponding to the selected arm, and updating an estimated value of the selected arm; updating the probability of each arm being selected according to a boltzmann distribution; and repeating the steps until the arms corresponding to all the normal sensor nodes are selected, and enabling the power of the signals sent by the sensor nodes to be optimal. The invention is suitable for the fault removal sensor.

Description

Elastic beamforming method based on multi-arm gambling machine in wireless sensor network

technical field

The invention belongs to the technical field of wireless communication networks, and in particular relates to an elastic beam forming method based on a multi-arm gambling machine in a wireless sensor network.

Background technique

In recent years, with the rapid development of sensor technology, communication technology, information processing technology and embedded technology, wireless sensor networks have gradually been widely used. Wireless sensor networks can perceive and collect information in the target area, and finally transmit the information to the user terminal. It has a wide range of application scenarios. Currently, it is mainly used in many fields such as area detection, medical monitoring, environmental monitoring, industrial monitoring, military defense and so on. .

The wireless sensor network is composed of a large number of cheap micro sensor nodes deployed in the monitoring area, and a multi-hop self-organizing network system formed by wireless communication. Its purpose is to sense, collect and process the sensed objects in the network coverage area through cooperation. information and send it to the observer. In reality, due to harsh and uncertain environments, battery power limitations, noise effects, and malicious attacks from adversaries, there are often some malicious sensors in the network to interfere with the system. If these malicious sensors are not processed in a timely manner, there is the potential for unreliable data, system instability, and loss of energy. Therefore, how to design a strategy to exclude malicious sensors, so as to ensure the correctness of all sensors that send signals, is an important issue faced by wireless sensor networks.

Beamforming technology is one of the technologies to improve the communication quality of wireless sensor network. Beamforming is a signal processing technique used in sensor arrays to achieve directional signal transmission or reception by causing signals at certain angles to undergo constructive interference while other signals undergo destructive interference, thereby enhancing the power of the transmitted signal. In the wireless sensor network, the signal sent by the node can be concentrated in the direction of the receiving end through the beamforming technology, so as to increase the signal power received at the receiving end. At the same time, the beamforming technology evenly distributes the energy consumption required for transmitting information among multiple nodes participating in the signal transmission, thus reducing the energy consumption of the nodes and enhancing the lifespan of the wireless sensor network. However, if a faulty sensor is involved in transmitting the signal, the signal power at the receiving end cannot be maximized, that is, beamforming cannot be achieved.

The Multi-Armed Bandit problem belongs to a field of reinforcement learning, the main idea of which is that each time the player chooses one of multiple given arms that allocate different resources according to a certain strategy, in order to find the best arm, Maximize the total reward over a series of trials. The beamforming problem with faulty sensors is transformed into a multi-arm gambling machine problem. By constructing an appropriate immediate reward, multiple selections are made and the probability of the selection arm is updated, so that the selection probability of the best arm is increased to 1, and finally each selection is made. The arm only contains normal sensor nodes, thus eliminating the influence of the faulty sensor and ensuring the maximization of the received signal power at the receiving end.

In the wireless sensor network, the accuracy of data collection is very high, and the sensor node needs to transmit the information to the user terminal accurately, so there cannot be the interference of the faulty sensor. In wireless sensor networks, a centralized approach was first proposed, involving a central node that receives all other node information and confirms its status by analyzing this information. Tang and Chow proposed an algorithm called Neighborhood Hidden Conditional Random Field (NHCRF), which uses received signal strength, frequency and signal delay to determine whether a node is healthy or not. This method can relax the independence assumption of hidden Markov models and can effectively provide reliable diagnostic results even under different networks. Chanak and Banerjee detect faulty sensor nodes by fault state, and propose a fuzzy rule-based classification and management method for faulty nodes. This scheme not only enhances the reusability of faulty nodes and overcomes the uncertainty problem, but also improves the overall performance of the network. To avoid the dependence on the central node, Lee and Choi proposed a distributed fault detection algorithm, in which each node judges the state of its neighbor nodes by comparing with its own state. The method relies on the agnostic diagnostic (AD) method of minimal domain knowledge to detect faulty nodes, and thus can be applied to various wireless sensor networks. AD utilizes a continuously updated correlation graph to describe the state of nodes, thereby gradually identifying faulty nodes.

Therefore, how to design a method to identify and avoid faulty sensors in wireless sensor networks is a problem worth studying.

SUMMARY OF THE INVENTION

Based on the above problems in the prior art, the present invention provides an elastic beam forming method based on a multi-arm gambling machine in a wireless sensor network.

In order to achieve the above-mentioned purpose of the invention, the present invention adopts the following technical solutions:

The elastic beamforming method based on dobby machine in wireless sensor network includes the following steps:

S1. There are m sensor nodes in the wireless sensor network; when there is interference from faulty sensor nodes, the set containing different sensors is defined as an arm, and there are K groups of arms, K=2 ^m -2; initialize the selected arm of each arm The times N _i =0, the estimated value of each arm Q _i (0)=1, and the probability of each arm being selected at the initial moment is p _i =1/K;

S2, randomly select an arm according to the probability of being selected, all sensor nodes in the selected arm perform a phase update according to the random grouping distributed beamforming method, increase the selected number of times corresponding to the selected arm by one, and update the corresponding to the selected arm. reward value, and update the estimated value of the selected arm;

S3. Update the probability of being selected for each arm according to the Boltzmann distribution;

S4. Steps S2 and S3 are repeated until the arms corresponding to all normal sensor nodes are selected, and the power of the signal sent by the sensor nodes is optimized.

As a preferred solution, all sensor nodes in the selected arm perform a phase update according to the random grouping distributed beamforming method, including the following steps:

S21. Initialize the phases of all sensor nodes in the selected arm;

S22, determine the grouping probability q, 0<q<1; randomly divide all sensor nodes into two groups G ₁ and G ₂ according to the grouping probability q;

S23. The sensor nodes in group G ₁ send signals to the receiving end in four time slots respectively, and perform phase shift according to the received feedback information; the sensor nodes in group G ₂ send signals to the receiving end in four time slots respectively. The signal is transmitted with zero phase offset within each slot.

As a preferred solution, performing the phase shift according to the received feedback information includes: after receiving the feedback in the first time slot, adjusting the phase to π; after receiving the feedback in the second time slot, adjusting the phase Adjusted to π/2; after receiving the feedback in the third time slot, the phase is adjusted to -ψ(n)-3π/2;

Where, ψ(n)=arctan((1+α(n))/(1-α(n))),

α(n)=[(P(4n+2)-P(4n))/(P(4n+2)-P(4n+1))],

P(4n), P(4n+1), P(4n+2) represent the signal power received by the receiving end of the first, second, and third time slots, respectively, where n=1; for the fourth time slot The phase offset of the slot, if the power value P(4n+3) received by the receiving end of the fourth time slot is greater than or equal to the power value P(4n) received at the initial moment of the iteration stage, then the phase offset is 0 ; otherwise, the phase offset is π.

As a preferred solution, the method for updating the reward value corresponding to the selected arm is:

Among them, P _i (old) is the signal power generated by the node contained in the selected arm sending a signal to the receiving end without any phase offset; P _i (new) is the signal generated by the node contained in the selected arm sending a signal to the receiving end for one phase update The signal power of ; T' is greater than 0 and is constant.

As a preferred solution, the method for updating the estimated value of the selected arm is:

Q _i (N _i )=Q _i (N _i -1)+β[R _i (N _i )-Q _i (N _i -1)], 0<β<1.

As a preferred solution, the method for updating the probability of each arm being selected is:

As a preferred solution, the method for optimizing the power of the signal sent by the sensor nodes in the step S4 is: randomly group the sensor nodes in the arms corresponding to all normal sensor nodes and adjust the phase according to the feedback of the receiving end for multiple times. Iterate until the signal power received by the receiver reaches the optimal value.

As a preferred solution, the method for optimizing the power of the signal sent by the sensor node in the step S4 specifically includes the following steps:

S41. Initialize the phase of each sensor node in the arm corresponding to all normal sensor nodes;

S42, determine the grouping probability q, 0<q<1; randomly divide all sensor nodes into two groups G ₁ and G ₂ according to the grouping probability q;

S43. Each iteration includes four time slots; the sensor nodes in group G ₁ send signals to the receiving end in each time slot, respectively, and perform phase shift according to the received feedback information; sensors in group G ₂ The node sends a signal to the receiver in each time slot, and the phase offset in each time slot is zero;

S44. Repeat steps S42 and S43 until the signal power received by the receiving end reaches an optimal value.

As a preferred solution, the sensor nodes are assigned to group G1 with probability q, and assigned to G2 group with probability _{1 -} q.

As a preferred solution, performing the phase shift according to the received feedback information includes: after receiving the feedback in the first time slot, adjusting the phase to π; after receiving the feedback in the second time slot, adjusting the phase Adjusted to π/2; after receiving the feedback in the third time slot, the phase is adjusted to -ψ(n)-3π/2;

Where, ψ(n)=arctan((1+α(n))/(1-α(n))),

α(n)=[(P(4n+2)-P(4n))/(P(4n+2)-P(4n+1))],

P(4n), P(4n+1), P(4n+2) represent the signal power received by the receivers of the first, second, and third time slots in the nth iteration, respectively; for the fourth time slot The phase offset of the slot, if the power value P(4n+3) received by the receiver in the fourth time slot is greater than or equal to the power value P(4n) received at the initial moment of its iteration stage, the phase offset is 0 ; otherwise, the phase offset is π.

Compared with the prior art, the present invention has the following beneficial effects:

The elastic beam forming method based on the multi-arm gambling machine in the wireless sensor network of the present invention takes the set of different sensors as one arm in the multi-arm gambling machine problem, and selects it according to a certain probability, using the idea of the multi-arm gambling machine problem Find the optimal arm, that is, all normal sensors, so as to avoid the influence of faulty sensors, and finally make the signals of all normal sensors perfectly coupled at the receiving end, so that the signal power sent by the sensor nodes is optimal. The method is suitable for the wireless sensor network to eliminate the faulty sensor in the system, improves the stability of the system, and can improve the signal transmission power.

Description of drawings

FIG. 1 is a flowchart of an elastic beam forming method based on a multi-arm gambling machine in a wireless sensor network according to an embodiment of the present invention;

2 is a coordinate diagram of the probability of normal sensor nodes being selected in the iterative process of the elastic beamforming method based on the multi-arm gambling machine in the wireless sensor network according to the embodiment of the present invention;

3 is a coordinate diagram of the optimal signal transmission power in the iterative process of the elastic beamforming method based on the multi-arm gambling machine in the wireless sensor network according to the embodiment of the present invention;

4 is a flowchart of the RG-DB algorithm in the elastic beamforming method based on the multi-armed gambling machine in the wireless sensor network according to the embodiment of the present invention;

5 is a schematic diagram of operations in each time slot in an iterative process of the RG-DB algorithm in the multi-arm gambling machine-based elastic beamforming method in the wireless sensor network according to the embodiment of the present invention.

Detailed ways

In order to describe the embodiments of the present invention more clearly, the specific embodiments of the present invention will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative efforts, and obtain other implementations.

The invention provides an elastic beam forming method based on a multi-arm gambling machine in a wireless sensor network, which aims to solve the problem that the signal power sent by the sensor node cannot reach the optimal value at the receiving end due to the existence of the faulty sensor. On the premise of low network energy consumption, a node strategy that can eliminate faulty sensors is found; a random grouping-based distributed beamforming method (or algorithm) (Random Grouping DistributedBeamforming algorithm, referred to as RG-DB algorithm) is also used to select sensors. The node performs phase adjustment, and finally the signal power sent by the sensor node can be maximized and received by the receiver.

The main idea of the elastic beam forming method based on the multi-arm gambling machine in the wireless sensor network of the embodiment of the present invention is to convert the problem of troubleshooting sensors into the problem of selecting the best arm for the multi-arm gambling machine. The phase of the selected arm (a group of sensor nodes) is adjusted by the RG-DB algorithm, and then the probability of the arm being selected is updated. Finally, all faulty sensor nodes can be eliminated, so that the signal power received by the receiver can be maximized. The above is the Multi-Armed Bandit based Resilient Beamforming algorithm (Multi-Armed Bandit based Resilient Beamforming), referred to as MAB-RB algorithm.

According to the application characteristics of this type of sensor network, the present invention establishes the following network model: in the wireless sensor network, there are m sensor nodes in total, and each sensor node can send a signal As(t) to the receiving end, where A represents The amplitude of the transmitted signal, s(t)=e ^iωt , i is the imaginary unit, and ω is the carrier frequency; each sensor node can also adjust its own phase offset according to the feedback information from the receiving end, and the transmitted signal spreads evenly in all directions .

It is assumed that each sensor node in the system has its own local oscillator, which can be synchronized to the carrier frequency ω. Divide the time into time periods with an iteration period of T, T=T _x + _Tr (T _x represents the time period during which the sensor sends signals, and _Tr represents the time period during which the node receives feedback information). In the RG-DB algorithm, each iteration includes four time slots, ie the nth iteration occurs during the time interval [4nT, 4nT+4T]. An iterative operation is performed, as shown in Figure 5. The signal will have a certain delay in the propagation process, and there will be a certain attenuation with the propagation distance.

The RG-DB algorithm is mainly used in wireless sensor networks. In each iteration cycle, the sensor nodes are randomly divided into two groups according to the grouping probability q, send signals in each time slot, and perform phase offset according to the feedback information sent by the receiver. shift. After many iterations, all signals are finally perfectly coupled at the receiving end, thereby maximizing the signal power at the receiving end. When a faulty sensor node interferes with the system, the set containing different sensors is defined as an arm, and there are K groups of arms in total, where K=2 ^m −2 (except for the case of all normal sensors or all faulty sensors). At this point, the elastic beamforming problem is transformed into a MAB problem, and the goal is to select the best arm that contains only normal sensors.

In order to adjust the phase offset of the sensor corresponding to the selected arm, a random group distributed beamforming method (RG-DB) algorithm is used. The following is a detailed step description of the RG-DB algorithm, as shown in Figure 4:

Step A, network initialization, that is, initialize the phase of each node in the network, and determine the grouping probability q, 0<q<1;

Step B: Divide all sensor nodes into two groups, each node is divided into group G ₁ with the probability of q, and the probability of 1-q is divided into group G ₂ ;

Step C: For each node in the group _G1 , send a signal to the receiving end according to the phase corresponding to each time slot, and receive feedback information. As shown in Figure 5, the specific phase adjustment strategy is: the phase offset in the first slot is π, the phase offset in the second slot is π/2, and the phase offset in the third slot is π/2. The phase offset of is -ψ(n)-3π/2;

Among them, the calculation method of ψ(n) is:

ψ(n)=arctan((1+a(n))/(1-α(n))),

a(n)=[(P(4n+2)-P(4n))/(P(4n+2)-P(4n+1))], (P(4n),P(4n+1), P(4n+2) represents the signal power received by the receiver in the first, second, and third time slots in the nth iteration, respectively). And, for the sensor nodes of group G ₁ , if the power value P(4n+3) received by the receiver of the third time slot is greater than or equal to the power value P(4n) received at the initial moment of the iteration stage, then the fourth The phase offset corresponding to the time slot is 0, otherwise it is π.

_In step D, the sensor nodes in group G2 send signals to the receiving end in each time slot respectively, and no phase update is performed in the four time slots, that is, the phase offset in each time slot is zero.

Step E: Repeat the above steps B, C, and D; after several cycles are performed, the signal power at the receiving end can reach the optimal value.

Through the above steps, the node phase offset can be adjusted, and finally the signal power at the receiving end can be maximized, thereby achieving optimal beamforming.

In addition, the MAB-RB algorithm mainly uses different sets of sensors as an arm in the multi-arm gambling problem, and selects it according to a certain probability, and uses the idea of the multi-arm gambling problem to find the optimal arm (that is, all normal sensors). ) to avoid the impact of faulty nodes.

Therefore, it is necessary to construct an appropriate immediate reward, and introduce the arm estimated value Q _i (N _i ) to evaluate the performance of arm i, where N _i is the number of times arm i is selected. Specifically, as shown in FIG. 1 , the elastic beam forming method based on the multi-arm gambling machine in the wireless sensor network of this embodiment includes the following steps:

Step 1: Network initialization. In a wireless sensor network including m sensor nodes, when there is interference from a faulty sensor node, there are K= ^2m -2 arms in total; the number of times each arm is selected for initialization is N _i =0 , the estimated value of each arm Q _i (0)=1, and the probability of each arm being selected is p _i =1/K;

Step 2, randomly select an arm according to the probability of being selected, all sensor nodes in the selected arm perform a phase update according to the random grouping distributed beamforming method, the number of selected arms corresponding to the selected arm is increased by 1, and the corresponding selected arm is updated. and update the estimated value of the selected arm; specifically, select an arm according to the probability, and all sensor nodes included in the arm perform a phase update according to the above-mentioned RG-DB algorithm (that is, the steps in the above-mentioned RG-DB algorithm A-D, also known as one iteration), and then the number of times the arm is selected is increased by one;

计算该臂对应的reward值，其中，P_i(old)是所选臂包含的节点没有任何相位偏移向接收端发送信号产生的信号功率；P_i(new)为所选臂包含的节点一次相位更新向接收端发送信号产生的信号功率；T′大于0，且为常数；use

Calculate the reward value corresponding to the arm, where P _i (old) is the signal power generated by the node contained in the selected arm sending a signal to the receiver without any phase offset; P _i (new) is the node contained in the selected arm once The signal power generated by the phase update sending the signal to the receiving end; T' is greater than 0 and is a constant;

The method to update the estimate for the selected arm is:

Q _i (N _i )=Q _i (N _i -1)+β[R _i (N _i )-Q _i (N _i -1)]; 0<β<1, which is a constant.

Step 3. For each arm (that is, all arms), update the probability of being selected according to the Boltzmann distribution. The method of updating the probability is:

为常数。

is a constant.

Step 4: Repeat steps 2 and 3. After several iterations, the arm containing all normal nodes can be selected, and the signal power sent by the sensor nodes can reach the optimal value.

Among them, the method for the sensor node to send the signal power to reach the optimal value includes:

S41. Initialize the phase of each sensor node in the arm corresponding to all normal sensor nodes;

S42. Determine the grouping probability q, 0<q<1; randomly divide all sensor nodes into two groups G ₁ and G ₂ according to the grouping probability q; the probability q of the sensor nodes being assigned to the G ₁ group, the probability of being assigned to the G ₂ group 1-q;

S43. Each iteration includes four time slots; the sensor nodes in group G ₁ send signals to the receiving end in each time slot, respectively, and perform phase shift according to the received feedback information; sensors in group G ₂ The node sends a signal to the receiving end in each time slot, and the phase offset in each time slot is zero; wherein, the phase offset is performed according to the received feedback information, including: receiving the first time After receiving the feedback in the slot, the phase is adjusted to π; after receiving the feedback in the second time slot, the phase is adjusted to π/2; after receiving the feedback in the third time slot, the phase is adjusted to -ψ(n) -3π/2;

Where, ψ(n)=arctan((1+a(n))/(1-a(n))),

a(n)=[(P(4n+2)-P(4n))/(P(4n+2)-P(4n+1))],

P(4n), P(4n+1), P(4n+2) represent the signal power received by the receivers of the first, second, and third time slots in the nth iteration, respectively; for the fourth time slot The phase offset of the slot, if the power value P(4n+3) received by the receiver in the fourth time slot is greater than or equal to the power value P(4n) received at the initial moment of its iteration stage, the phase offset is 0 ; otherwise, the phase offset is π.

S44. Repeat steps S42 and S43 until the signal power received by the receiving end reaches an optimal value.

The specific application example of the elastic beam forming method based on the multi-arm gambling machine in the wireless sensor network of the present invention is as follows:

1. Network topology and parameter setting, set 8 sensors, including 6 normal sensors and 2 faulty sensors, the phase of the faulty sensor always changes with time and is uncontrollable; all sensors have the same attenuation amplitude υ _ι A =0.5, grouping probability q=0.4; to simplify the embodiment, assuming that the number of faulty sensors is known, then the number of arms

2. Initialize the estimated value of each arm Q _i (0)=1, set the number of times each arm is selected N _i =0, and the probability of each arm being selected is p _i =1/K;

3. An arm is randomly selected according to the probability, the nodes contained in the arm perform one iteration according to the RG-DB algorithm, and the number of selected arms corresponding to the arm is increased by one; and the reward, estimated value and selected arm corresponding to the arm are updated according to the corresponding formula. The probability;

4. Repeat step 3 continuously, and finally it can be obtained that the probability of all correct sensors being selected tends to 1, as shown in Figure 2, and the signal power values of all normal sensors converge to the maximum value of 9dB, as shown in Figure 3.

From this, it can be obtained that all normal sensor nodes, that is, the optimal arm, are finally selected each time. At this time, the faulty sensor is eliminated, and the signal transmission power reaches the maximum.

The above is only a detailed description of the preferred embodiments and principles of the present invention. For those of ordinary skill in the art, according to the ideas provided by the present invention, there will be changes in the specific implementation, and these changes should also be It is regarded as the protection scope of the present invention.

Claims (10)

1. The elastic beam forming method based on the multi-arm gambling machine in the wireless sensor network, is characterized in that, comprises the following steps: S1. There are m sensor nodes in the wireless sensor network; when there is interference from faulty sensor nodes, the set containing different sensors is defined as an arm, and there are K groups of arms, K=2 ^m -2; initialize the selected arm of each arm The times N _i =0, the estimated value of each arm Q _i (0)=1, and the probability of each arm being selected at the initial moment is p _i =1/K; S2, randomly select an arm according to the probability of being selected, all sensor nodes in the selected arm perform a phase update according to the random grouping distributed beamforming method, increase the selected number of times corresponding to the selected arm by one, and update the corresponding to the selected arm. reward value, and update the estimated value of the selected arm; S3. Update the probability of being selected for each arm according to the Boltzmann distribution; S4. Steps S2 and S3 are repeated until the arms corresponding to all normal sensor nodes are selected, and the power of the signal sent by the sensor nodes is optimized. 2. The elastic beamforming method based on a multi-arm gambling machine in the wireless sensor network according to claim 1, wherein all sensor nodes in the selected arms perform a phase update according to the random grouping distributed beamforming method , including the following steps: S21. Initialize the phases of all sensor nodes in the selected arm; S22, determine the grouping probability q, 0<q<1; randomly divide all sensor nodes into two groups G ₁ and G ₂ according to the grouping probability q; S23. The sensor nodes in group G ₁ send signals to the receiving end in four time slots respectively, and perform phase shift according to the received feedback information; the sensor nodes in group G ₂ send signals to the receiving end in four time slots respectively. The signal is transmitted with zero phase offset within each slot. 3. The elastic beamforming method based on a multi-armed gambling machine in a wireless sensor network according to claim 2, wherein the performing phase shift according to the received feedback information comprises: receiving the first time After receiving the feedback in the slot, the phase is adjusted to π; after receiving the feedback in the second time slot, the phase is adjusted to π/2; after receiving the feedback in the third time slot, the phase is adjusted to -ψ(n) -3π/2; Where, ψ(n)=arctan((1+α(n))/(1-α(n))), α(n)=[(P(4n+2)-P(4n))/(P(4n+2)-P(4n+1))], P(4n), P(4n+1), P(4n+2) represent the signal power received by the receiving end of the first, second, and third time slots, respectively, where n=1; for the fourth time slot The phase offset of the slot, if the power value P(4n+3) received by the receiving end of the fourth time slot is greater than or equal to the power value P(4n) received at the initial moment of the iteration stage, then the phase offset is 0 ; otherwise, the phase offset is π. 4. The elastic beamforming method based on a multi-arm gambling machine in the wireless sensor network according to claim 3, wherein the method for updating the reward value corresponding to the selected arm is:

Among them, P _i (old) is the signal power generated by the node contained in the selected arm sending a signal to the receiving end without any phase offset; P _i (new) is the signal generated by the node contained in the selected arm sending a signal to the receiving end for one phase update The signal power of ; T' is greater than 0 and is constant. 5. The elastic beamforming method based on a multi-arm gambling machine in the wireless sensor network according to claim 4, wherein the method for updating the estimated value of the selected arm is: Q _i (N _i )=Q _i (N _i -1)+β[R _i (N _i )-Q _i (N _i -1)], 0<β<1. 6. The elastic beamforming method based on multi-arm gambling machine in the wireless sensor network according to claim 5, is characterized in that, the method that updates the probability that each arm is selected is:

7. The elastic beamforming method based on a multi-armed gambling machine in the wireless sensor network according to any one of claims 1-6, wherein the step S4 is to make the power of the signal sent by the sensor node to reach an optimal level. The method is as follows: randomly group the sensor nodes in the arms corresponding to all normal sensor nodes and adjust the phase according to the feedback of the receiver to perform multiple iterations until the signal power received by the receiver reaches the optimal value. 8. The elastic beam forming method based on a multi-armed gambling machine in a wireless sensor network according to claim 7, wherein the method for optimizing the power of the signal sent by the sensor node in the step S4 specifically comprises the following steps: step: S41. Initialize the phase of each sensor node in the arm corresponding to all normal sensor nodes; S42, determine the grouping probability q, 0<q<1; randomly divide all sensor nodes into two groups G ₁ and G ₂ according to the grouping probability q; S43. Each iteration includes four time slots; the sensor nodes in group G ₁ send signals to the receiving end in each time slot, respectively, and perform phase shift according to the received feedback information; sensors in group G ₂ The node sends a signal to the receiver in each time slot, and the phase offset in each time slot is zero; S44. Repeat steps S42 and S43 until the signal power received by the receiving end reaches an optimal value. 9 . The elastic beamforming method based on a multi-armed gambling machine in a wireless sensor network according to claim 8 , wherein the sensor nodes are assigned to the probability q of group G ₁ , and the probability of being assigned to group G ₂ is 1- q. 10 . The elastic beamforming method based on a multi-armed gambling machine in a wireless sensor network according to claim 8 , wherein the performing phase shift according to the received feedback information comprises: receiving the first time After receiving the feedback in the slot, the phase is adjusted to π; after receiving the feedback in the second time slot, the phase is adjusted to π/2; after receiving the feedback in the third time slot, the phase is adjusted to -ψ(n) -3π/2; Where, ψ(n)=arctan((1+α(n))/(1-α(n))), α(n)=[(P(4n+2)-P(4n))/(P(4n+2)-P(4n+1))], P(4n), P(4n+1), P(4n+2) represent the signal power received by the receivers of the first, second, and third time slots in the nth iteration, respectively; for the fourth time slot The phase offset of the slot, if the power value P(4n+3) received by the receiver in the fourth time slot is greater than or equal to the power value P(4n) received at the initial moment of the iteration stage, then the phase offset is 0 ; otherwise, the phase offset is π.

Images