CN102186241A - Parallel distributed particle filter based wireless sensor network target tracking method - Google Patents

Abstract

The invention discloses a parallel distributed particle filter based wireless sensor network target tracking method which comprises the following steps of: replacing a particle and a weight by adopting a Gaussian mixture model, carrying out parallel computation on local importance sampling, local weights, the sum of the local weights, local state estimation and variance of each node, and transmitting data to a fusion center; computing the general state estimation according to the transmitted data to obtain a state estimation value and a state estimation variance at the current moment by the fusion center; and transmitting the state estimation value and the state estimation variance computed at the previous moment as well as the particle and the weight replaced by the Gaussian mixture model to a clustering node at the current moment, and performing the state estimation on a target position. In the invention, the Gaussian mixture model is only transmitted between cluster heads, and the particle and the weight replaced by the Gaussian mixture model only need to be transmitted between the clusters when the cluster heads are replaced, thus, the computing quantity of an algorithm is reduced, the bandwidth is further reduced effectively, and the real time and the tracking accuracy are improved.

Description

Wireless sensor network target tracking method based on parallel distributed particle filtering

Technical Field

The invention relates to a target tracking method of a wireless sensor network, belonging to the technical field of sensor technology and wireless communication.

Background

The wireless sensor network is composed of a large number of randomly distributed sensor nodes, is a self-organizing network and can sense the information of detected objects in a coverage area. The sensor node has the characteristics of small volume, low price, wireless communication, self-organization, robustness, concealment and the like, and is widely applied to the fields of national defense and military, environmental detection and the like. One of typical applications of the wireless sensor network is target tracking, which tracks a target entering a coverage area through mutual cooperation among nodes, and estimates information such as a position of the target by using a multi-node observation value and filtering. However, wireless sensor networks also have their own drawbacks: 1) the energy is limited, the general sensor nodes are all powered by batteries and cannot be recharged, and the transmission of information consumes a large amount of energy; 2) the bandwidth is limited, and a large amount of data needs to be transmitted in the cooperative work among the nodes; 3) the nodes have limited computational power. At present, most target tracking filtering based on a wireless sensor network adopts an angle or signal strength as a measurement value, which causes excessive data transmission of a sensor, data congestion and occupation of a large amount of bandwidth, so that the traditional target tracking is difficult to be applied to a common wireless sensor network.

Compared with the common wireless sensor network, the binary wireless sensor network only transmits '0' or '1', so that the bandwidth and the energy can be effectively saved. The existing binary wireless sensor network target tracking uses network structure, sensor detection radius and geometric knowledge to carry out positioning, and is deficient in precision; or a filtering method is used for tracking the target, but most of the methods are centralized filtering algorithms, so that too much energy is consumed and the real-time performance cannot meet the requirements.

At present, a target tracking method of a distributed particle filter wireless sensor network can perform dynamic clustering according to the current target position, and is optimized in reducing energy consumption and bandwidth, but the defects are as follows: firstly, the particle filtering algorithm in the method needs a large amount of time, and the real-time performance is not good; secondly, the method is to calculate on the cluster head node, which can lead the cluster head node to die prematurely.

Disclosure of Invention

The invention aims to overcome the defects of a target tracking method of a distributed particle filter wireless sensor network in the prior art, provides a binary system wireless sensor network target tracking method based on parallel distributed particle filters, and averagely loads the calculated amount of cluster head nodes by fully utilizing the distributed characteristics so as to balance energy consumption.

The technical scheme of the invention comprises the following steps: 1) randomly broadcasting wireless sensor network nodes in a coverage area, selecting cluster head nodes according to a cluster principle, enabling the cluster nodes to observe targets, and sending binary data to the cluster head nodes; 2) forming an initial cluster at the time when t is 0, selecting cluster head nodes, firstly sampling from the prior distribution, adopting a Gaussian mixture model to replace particles and weights, and enabling all particles X to be_tDividing the data into s subsets, dividing the subsets into each node and updating the particles; local importance samples are then computed in parallel for each node

Local weight

Local weight sum

Local state estimation

Sum variance

Will obtain

Transmitting to the fusion center and resampling locally; finally, the fusion center calculates the overall state estimation according to the transmitted data to obtain the state estimation value of the current momentAnd the state estimation variance P_t(ii) a 3) Grouping clusters at the time t according to a cluster grouping principle, and transmitting the state estimation value and the state estimation variance value calculated at the previous time, and the particles and the weight replaced by the Gaussian mixture model to the cluster head node at the time; 4) calculating a state estimation value and a state estimation variance at the time t, and performing state estimation on a target position; 5) adding 1 at the time t, and repeating the steps 3) -4) until the target leaves the coverage area.

The invention has the beneficial effects that:

1. by adopting the binary wireless sensor network, only 0 or 1 signals are transmitted between the nodes in the cluster and the cluster head node, so that the bandwidth can be effectively reduced, and the energy consumption is greatly reduced.

2. The local parallel distributed particle filtering algorithm adopting the Gaussian mixture model is adopted, only the Gaussian mixture model is transmitted between the cluster heads, the cluster and the cluster heads are continuously updated according to the movement of the target, when the cluster heads are replaced, only the particles and the weight replaced by the Gaussian mixture model need to be transmitted between the cluster heads, a large number of particles do not need to be transmitted between the cluster heads, and the bandwidth is further effectively reduced.

3. The parallel particle filtering algorithm is adopted, the particle number can be adjusted on line according to the filtering variance, the calculated amount of the algorithm is reduced, the communication traffic in the target tracking process is effectively reduced, the real-time performance and the tracking precision are greatly improved, and the energy consumption is reduced.

Drawings

The invention will be further described in detail with reference to the following drawings and detailed description:

FIG. 1 is a flow chart of a target tracking method of the present invention;

fig. 2 is a flow chart of local distributed particle filtering.

Detailed Description

The invention relates to a parallel particle filtering binary wireless sensor network target tracking method based on a Gaussian mixture model, wherein the Gaussian mixture model is a model formed by accurately quantizing things by using a Gaussian probability density function (normal distribution curve) and decomposing one thing into a plurality of Gaussian probability density functions. The method adopts parallel particle filtering to track the position of a target, and awakens part of sensor nodes to participate in target tracking according to a certain principle at the current moment so as to reduce communication traffic and energy consumption. Firstly, the wireless sensor nodes are in a dormant state, after a target enters the sensor nodes, the nodes are waken up, cluster nodes in a specified range are clustered, then cluster head nodes are selected, and cluster nodes which do not conform to the cluster principle enter the dormant state again, so that the energy of the sensor nodes is saved. The target tracking node uses a parallel particle filter algorithm, the cluster head node and the nodes in the cluster cooperate to use the parallel particle filter to estimate the target position at the current moment, the cluster and the cluster head node are continuously updated according to the movement of the target, and when the cluster head node is replaced, the information of the previous cluster head node is transmitted to the current cluster head node. Because the sensor nodes transmit binary data to the cluster heads and the particle number can be adjusted online, the bandwidth and the energy can be further saved.

Referring to fig. 1, the specific implementation method of the present invention is as follows:

step 101: and at the moment t being 0, initializing the wireless sensor network, randomly broadcasting nodes of the wireless sensor network in a coverage area, wherein all the nodes have uniform specifications such as communication distance, detection distance and the like, all the nodes are in a dormant state, only a simple detection function is kept, namely only the existence or nonexistence of a target can be detected, and the communication function is closed.

Step 102: the sensor nodes detect the target and wake up the nodes in the detection range, and in the nodes, the nodes are grouped according to a principle, wherein the grouping principle is as follows: selecting a node k with the maximum signal receiving intensity as a cluster head node, clustering the node k with the cluster head node in a single-hop communication range and the cluster head node, wherein the node in the cluster is in an awakening state, and the rest nodes continue to enter a dormant state;

step 103: and observing the target by the nodes in the cluster, and sending binary data to the cluster head node.

The intensity model of the received signal when the node in the cluster observes the target is as follows:

n＝1，2，L，M，g_n(x_t) Is a function of the received signal strength of the nth node, v_n，tIs a noise, and the noise is,

μ_vis v_n，tThe average value of (a) of (b),

is v_n，tVariance of r_nIs the position of the nth node, I_tIs the position of the target at time t, | | r_n-I_tI | is the Euclidean distance between the target and the node, Ψ is the distance d at the target₀The signal energy of time, α, is a parameter related to the transmission medium.

The signal strength y received by the nth node_n，tProcessing locally, the received signal y_n，tComparing with a preset threshold gamma value, and if the value is lower than the threshold gamma value, not sending any information; if it is above the threshold value of gamma,binary information is sent to the cluster head node.

The cluster head node receives the measurement z from the nth node_n，t＝β_ns_n，t+ε_n，tWherein

ε_n，tIt is the observation of the noise that is, is the variance, beta_nIs a parameter related to the kind of sensor.

Step 104: at the moment t is 0, forming an initial cluster according to the cluster forming principle of the step 102, selecting a cluster head node, and performing the prior distribution p (x)₀) I.e. sampling in the initial state

And (i is 1, 2, L, N (N is the number of particles), establishing a Gaussian mixture model by adopting a conventional method, replacing particles and weights with the Gaussian mixture model, and estimating initial state estimation and variance estimation by using a parallel particle filter algorithm. The method comprises the following specific steps:

first, the particles are extracted with a weight of

And:i represents a particle, t represents a time,

i particle at time t, z_tFor the total measurement at time t, p (-) is a probability density function because of the observed noise ε_n，tAre independent, so:

thus, it is possible to provideCan be written as:

$p\left(z_{n,t}\right|x_t^i)=p\left(z_{n,t}\right|s_{n,t}=0,x_t^i)p(s_{n,t}=0|x_t^i)+p\left(z_{n,t}\right|s_{n,t}=1,x_t^i)p(s_{n,t}=1|x_t^i)$

$=p\left(z_{n,t}\right|s_{n,t}=0)p(s_{n,t}=0|x_t^i)+p\left(z_{n,t}\right|s_{n,t}=1)p(s_{n,t}=1|x_t^i),$

wherein z is_n，tIs the measurement of the nth node at time t,

and is

Where Q (-) is a cumulative function of normal distribution.

Then, referring to fig. 2, the parallel algorithm is performed as follows:

1) the fusion center is the entire particle X_tInto subsets of particles

s＝1，L，S：

with

Means at time t, at the s-th node, m_sParticles of N_sIs the number of particles that are sorted over s nodes. And distributing the particle subset to each node and updating the particles.

2) The fusion center is corresponding to each node s as 1, LS parallel computation of local importance samples

Local weight

Local weight sum

Normalizing local weight to

Local state estimation

And local varianceWill obtain

Transmitting to the fusion center; and resamples the local.

3) After the fusion center receives the data transmitted by the nodes, the overall state estimation is calculated to obtain the state estimation value at the current moment

Meanwhile, the total state estimation variance is calculated to obtain the state estimation variance P at the current moment_t。

The local weight sum mentioned above

Local state estimation

Local variance

Current time state estimation

The state estimation variance P at the current time_tCalculated by the following equation:

order to

For the sampled particles and non-normalized weights for node s, then the local estimate and variance given at node s is:

wherein

To normalize the weights, the following formula is calculated:

to combine all particles without normalization, the global estimates and variances given at the fusion center are:

therein

For global normalization weight, the following formula is used for calculation:

from the several equations given above, it can be seen that the global state estimate and variance can be calculated from the local state estimate and variance as follows:

<math><mrow><msub><mi>P</mi><mi>t</mi></msub><mo>=</mo><munderover><mi>&Sigma;</mi><mrow><mi>s</mi><mo>=</mo><mn>1</mn></mrow><mi>S</mi></munderover><msubsup><mi>w</mi><mi>t</mi><mi>s</mi></msubsup><mo>[</mo><msubsup><mi>P</mi><mi>t</mi><mi>s</mi></msubsup><mo>+</mo><mrow><mo>(</mo><msubsup><mover><mi>X</mi><mo>^</mo></mover><mi>t</mi><mi>s</mi></msubsup><mo>-</mo><msub><mover><mi>X</mi><mo>^</mo></mover><mi>t</mi></msub><mo>)</mo></mrow><mrow><mo>(</mo><msubsup><mover><mi>X</mi><mo>^</mo></mover><mi>t</mi><mi>s</mi></msubsup><mo>-</mo><msub><mover><mi>X</mi><mo>^</mo></mover><mi>t</mi></msub><mo>)</mo></mrow><mo>&prime;</mo><mo>]</mo><mo>,</mo></mrow></math> wherein,

step 105: at the moment t, clustering according to a clustering principle, packing and transmitting a state estimation value and a variance value calculated by particle filtering at the previous moment, particles replaced by a low-dimensional Gaussian mixture model and weight to a cluster head node at the moment;

step 106: and (5) performing a parallel particle filter algorithm at the time t, wherein the parallel algorithm performs state estimation on the position of the target in step 104.

Step 107: adding 1 at any moment;

step 108: step 105 and step 107 are repeated until the target leaves the coverage area.

Claims (3)

1. A wireless sensor network target tracking method based on parallel distributed particle filtering is characterized by comprising the following steps:

1) randomly broadcasting wireless sensor network nodes in a coverage area, selecting cluster head nodes according to a cluster principle, enabling the cluster nodes to observe targets, and sending binary data to the cluster head nodes;

2) forming an initial cluster at the time when t is 0, selecting cluster head nodes, firstly sampling from the prior distribution, adopting a Gaussian mixture model to replace particles and weights, and enabling all particles X to be_tIs divided into s piecesThe subset is distributed to each node and particle updating is carried out; local importance samples are then computed in parallel for each nodeLocal weight

Local weight sum

Local state estimationSum varianceWill obtainTransmitting to the fusion center and resampling locally; finally, the fusion center calculates the overall state estimation according to the transmitted data to obtain the state estimation value of the current momentAnd the state estimation variance P_t；

3) Grouping clusters at the time t according to a cluster grouping principle, and transmitting the state estimation value and the state estimation variance value calculated at the previous time, and the particles and the weight replaced by the Gaussian mixture model to the cluster head node at the time;

4) calculating a state estimation value and a state estimation variance at the time t, and performing state estimation on a target position;

5) adding 1 at the time t, and repeating the steps 3) -4) until the target leaves the coverage area.

2. The method for tracking the target of the wireless sensor network based on the parallel distributed particle filtering as claimed in claim 1Characterized in that: step 1) the intensity model of the received signal when the node in the cluster observes the target is as follows:n＝1，2，L，M，g_n(x_t) Is a function of the received signal strength of the nth node, v_n，tIs a noise, and the noise is,

μ_vis v_n，tThe average value of (a) of (b),

is v_n，tVariance of r_nIs the position of the nth node, I_tIs the position of the target at time t, | | r_n-I_tI | is the Euclidean distance between the target and the node, Ψ is the distance d at the target₀The signal energy of time, α, is a parameter related to the transmission medium.

3. The method for tracking the target of the wireless sensor network based on the parallel distributed particle filtering as claimed in claim 1, wherein: the global state estimate and variance are calculated from the local state estimate and variance, the calculation formula is:

wherein,

Images