WO2023221656A1 - Information fusion-based wireless sensor network positioning method for marine search and rescue - Google Patents

Abstract

Provided in the present invention is an information fusion-based wireless sensor network positioning method for marine search and rescue. The method includes: S1, respectively constructing an RSS-based ranging model and a TOA-based ranging model in combination with sea wave shielding noise; S2, according to constraints, fusing multi-source ranging information in the RSS-based ranging model and the TOA-based ranging model, and constructing a hybrid ranging constrained least squares framework; S3, introducing a buffer factor, and by using a method for improving the rotation of block principal elements, acquiring an initial solution of a target position by means of cyclic alternation of feasible solution indexes; and S4, on the basis of a first-order Taylor series expansion, deriving a linear re-optimization method, and performing error correction on the initial solution, which is obtained in S3, so as to obtain a more accurate solution for the position. The method has the advantage that the problem of the positioning accuracy being reduced due to an increase in the area of a monitoring region and in noise resulting from the use of marine wireless sensor network positioning technology, which only relies on a single ranging means, is solved.

Description

An information fusion wireless sensor network positioning method for maritime search and rescue Technical field

The invention relates to the technical field of marine wireless sensor network node positioning, and in particular to an information fusion marine search and rescue wireless sensor network positioning method.

Background technique

As an important part of the maritime economy, the development of the maritime transportation industry is crucial. In order to ensure the safety of maritime transportation, relevant departments have formulated a series of measures to prevent accidents, especially those caused by human factors. However, the maritime environment is complex and changeable, and the occurrence of extreme weather threatens the safety of maritime transportation, such as the Yangtze Star cruise ship accident. When extreme weather causes a maritime accident, how to reduce the loss of life and property as much as possible is a very critical link. As the last barrier to ensure the safety of life at sea, Marine Search and Rescue (MSR) can reduce the loss of life and property to the greatest extent through multi-faceted and multi-department integrated cooperation. At this stage, maritime search and rescue mainly uses remote sensing images or based on wind currents and other speeds to narrow the search and rescue scope as much as possible to facilitate rescue operations. However, this method is time-consuming and has large errors, and may miss the golden period of rescue to a large extent. In order to improve this shortcoming, the good self-organization, scalability and adaptability of Wireless Sensor Networks (WSNs) can be used to greatly improve the search and rescue success rate and efficiency.

However, how to accurately and efficiently locate rescue targets in maritime search and rescue sensor networks is a challenge. On the one hand, in the highly dynamic environment at sea, rescue targets will move with wind currents, etc., making positioning difficult and inefficient; on the other hand, high-latency, low-bandwidth communication channels at sea make positioning accuracy poor, coupled with wave shadowing The non-linear non-Gaussian noise generated by multipath effect and other effects further increases the positioning error. In addition, relying only on a single ranging technology, such as Received Signal Strength (RSS), the error will increase as the search and rescue range increases due to inherent flaws in the technology. The existing positioning technology does not have a good solution to the above three problems of maritime search and rescue wireless sensor networks. It cannot take into account the accuracy and efficiency of positioning well, and cannot achieve real-time, efficient and accurate positioning of targets to be rescued.

Contents of the invention

The purpose of the present invention is to provide an information fusion maritime search and rescue wireless sensor network positioning method to solve the problem of the increasing monitoring area area of the maritime wireless sensor network positioning technology that only relies on a single ranging method. The problem of reduced positioning accuracy caused by increased noise. This method is suitable for highly dynamic marine environments and can maintain good positioning performance in a larger monitoring area and under higher wave masking noise.

In order to achieve the above objects, the present invention is achieved through the following technical solutions:

A maritime search and rescue wireless sensor network positioning method based on information fusion, which is characterized by including the following steps:

S1. Considering the wave shadowing noise, construct the RSS and TOA ranging models respectively;

S2. According to the constraints, integrate RSS and TOA multi-source ranging information to construct a hybrid ranging constrained least squares framework;

S3. Introduce a buffer factor and use a method to improve block pivot rotation to obtain the initial position of the target. Specifically: divide the variable index to be found into two sets and define a subset of the corresponding variables, calculate its complementary basis solution and judge Whether it is a feasible solution, if not, the set is updated, and the update process introduces a buffer factor; the above judgment and update steps are alternately cyclically alternated, and finally all solutions to the variables to be found are feasible solutions, that is, the initial solution of the target position;

S4. Based on the first-order expansion of Taylor series, a linear re-optimization method is derived to correct the error of the initial solution obtained in S3, thereby obtaining a more accurate solution.

The RSS and TOA ranging models are constructed as described in step S1, specifically including:

S11. If there are N anchor nodes in the network, the position of the i-th anchor node at time t can be expressed as where T stands for transpose. The position of the target to be rescued at time t is expressed as The anchor node can receive signal strength information propagated from the target through radio signals at each moment, namely:

A ranging model based on the signal strength information propagated by radio signals is constructed, that is, the RSS ranging model; where, Indicates the transmit power value of the target received by the i-th anchor node at time t; d ₀ indicates the reference distance value, usually 1m; represents the target’s transmit power at time t; PL(d ₀ ) represents the loss value of the transmitted signal strength at the relevant reference distance; α ^t represents the path loss factor at time t; Represents the second-order norm; Indicates that the mean value in the RSS ranging model is zero and the variance is Wave obscuring noise satisfying Gaussian distribution.

S12. Similarly, the anchor node can also receive arrival time information from the target, which can be expressed as:

A model that measures distance based on signal propagation time is constructed, namely the TOA model; where Indicates that the mean value in the TOA ranging model is zero and the variance is Wave obscuring noise satisfying Gaussian distribution.

As described in step S2, the multi-source ranging information in the RSS ranging model and the TOA ranging model is integrated to construct a hybrid ranging constrained least squares framework, which specifically includes:

S21. According to the expression (1a) obtained in S11, we can obtain:

in,

S22. Linearly expand the equation (2) obtained in S21 to get:

S23. Consider constraints as well as Then the original positioning problem can be constructed as an unconstrained least squares framework expression based on hybrid ranging, that is:

S24. Perform square expansion on equation (4) obtained in S23 to get:

S25, another As a variable, where M represents the quantity to be found in the variable, combined with the constraint θ ^t ≥ 0, the original positioning problem can be further constructed as a hybrid ranging constrained least squares (HM-CLS) framework, that is:

in,

The buffer factor is introduced as described in step S3, and a method of block pivot rotation is used to obtain the initial solution of the target position through cyclically alternating feasible solution indexes, which specifically includes:

S31. Divide the variable index to be found into two sets, namely κ and in In addition, define ξ _κ , Ψ _κ , is a subset of the corresponding variables, and is the submatrix of the corresponding matrix A ^t .

S32. Calculate the complementary basis solution according to formula (8) and formula (9) If the constraint ξ _κ ≥ 0 is satisfied and Then it is a feasible solution, otherwise it is an infeasible solution.

S33. When the complementary base solution does not satisfy the constraints, define the set Γ to satisfy:

S34. For j∈Γ, variable ξ _j is an infeasible solution. Further update κ and

Among them, R is a non-empty subset and

S35. When the update rule falls into a loop or cannot find the corresponding solution, use equation (12) as a candidate update rule to find a feasible solution, that is:
R＝{j:j＝max{j∈Γ}} (12)

S36. In the update process, introduce a buffer factor When the infeasible solutions increase, the buffer is due to The child decreases. If the buffer factor is 0, the candidate update rule is used to find a feasible solution and stored in β.

S37: Perform cyclic updates based on S32 to S36 until all solutions to the variables to be obtained are feasible solutions.

Based on the first-order expansion of the Taylor series described in step S4, a linear re-optimization method is derived to correct the error of the initial solution obtained in S3, thereby obtaining a more accurate position solution. Specifically includes:

S41. Initial solution obtained according to S36 Further construct the loss function, namely:

S42. Use the first-order Taylor series expansion formula to exist near expand, where Represents the first two terms of variable θ ^t ; represents the final estimate. Then the corresponding expression can become:

in,

S43. Combine steps S41 and S42 to obtain the function. The loss function can be further expressed as:

S44, get the expression through S43, for To find the partial derivative, then:

S45. Let the expression obtained in S44 be 0, and the error can be obtained. Then perform the error correction operation to obtain:

Compared with the existing technology, the present invention has the following advantages: the method is adaptable to highly dynamic marine environments, and can maintain better positioning performance in a larger range of monitoring areas and under higher wave shielding noise, solving the problem of relying solely on Maritime wireless sensor network positioning technology with a single ranging method suffers from the problem of reduced positioning accuracy due to the increase in monitoring area and noise.

Description of the drawings

Figure 1 shows the structure diagram of the maritime search and rescue wireless sensor network system.

Figure 2 is a flow chart of an information fusion maritime search and rescue sensor network positioning method according to the present invention.

Figure 3 is the pseudo code of the improved block pivot rotation positioning method of the present invention.

Figure 4(a) and Figure 4(b) show the positioning performance of different monitoring areas of the present invention.

Figure 5(a) and Figure 5(b) show the positioning performance of different noises of the present invention.

Detailed ways

The present invention will be further elaborated below by describing in detail a preferred specific implementation case in conjunction with the accompanying drawings.

Figure 1 shows the structural block diagram of the maritime search and rescue sensor network. When a ship is in distress, the target to be rescued (the dot with the larger diameter in Figure 1) carries a life jacket with a relevant node device, and the rescue helicopter goes to the relevant sea area to disperse a GPS device. Or the Beidou signal, that is, the anchor node (the smaller diameter circle in Figure 1) that can obtain the position in real time. The anchor node and the node on the target to be rescued (target node) form a wireless sensor network through the Zigbee protocol, and then search for the rescue target. The problem becomes one of locating network nodes. After obtaining the location of the target to be rescued, the anchor node transmits the relevant information to the rescue ship and satellite, and the satellite transmits it to the relevant departments on land through signals and networks to implement specific rescue plans.

Assuming that there are N anchor nodes in the network, the position of the i-th anchor node at time t can be expressed as The position of the target to be rescued at time t is expressed as The anchor node can receive the Time of Arrival (TOA) and Signal Strength (RSS) information propagated by radio signals from the target at each moment.

As shown in Figure 2, in order to locate the target, the present invention proposes an information fusion maritime search and rescue wireless sensor network positioning method (Lightweight Computational Localization Technology using Information Fusing, LCCT-IF), which specifically includes:

S1. Considering the wave shadowing noise, construct the RSS and TOA ranging models respectively;

S2. According to the constraints, integrate the multi-source ranging information in the RSS ranging model and the TOA ranging model, and construct a hybrid ranging constrained least squares framework;

S3. Introduce a buffer factor, use a method to improve block pivot rotation, and obtain the initial solution of the target position through cyclically alternating feasible solution indexes;

S4. Based on the first-order expansion of Taylor series, a linear re-optimization method is derived to correct the error of the initial solution obtained in S3, thereby obtaining a more accurate solution.

In this implementation case, the step S1 specifically includes:

S11. The received RSS information, that is, the RSS ranging model can be expressed as:

in, Indicates the transmit power value of the target received by the i-th anchor node at time t; d ₀ indicates the reference distance value, usually 1m; represents the target’s transmit power at time t; PL(d ₀ ) represents the loss value of the transmitted signal strength at the relevant reference distance; α ^t represents the path loss factor at time t; Represents the second-order norm; Indicates that the mean of the RSS ranging model is zero and the variance is Wave obscuring noise satisfying Gaussian distribution.

S12. Similarly, the anchor node can also receive Time of Arrival (TOA) information from the target, that is, the TOA ranging model can be expressed as:

in Indicates that the mean value of the TOA ranging model is zero and the variance is respectively Wave obscuring noise satisfying Gaussian distribution.

The step S2 specifically includes:

S21. According to the expression (1a) obtained in S11, we can obtain:

in,

S22. Linearly expand the equation (2) obtained in S21 to get:

S23. Consider constraints as well as Then the original positioning problem can be constructed as an unconstrained least squares framework expression based on hybrid ranging, that is:

S24. Perform square expansion on equation (4) obtained in S23 to get:

S25, another As a variable, where M represents the quantity to be found in the variable, combined with the constraint θ ^t ≥ 0, the original positioning problem can be further constructed as a hybrid ranging constrained least squares (HM-CLS) framework, that is:

in,

The step S3 specifically includes:

S31. Divide the variable index to be found into two sets, namely κ and in In addition, define ξ _κ , Ψ _κ , is a subset of the corresponding variables, and is the submatrix of the corresponding matrix A ^t .

S32. Calculate the complementary basis solution according to formula (8) and formula (9) If the constraint ξ _κ ≥ 0 is satisfied and Then it is a feasible solution, otherwise it is an infeasible solution.

S33. When the complementary base solution does not satisfy the constraints, define the set Γ to satisfy:

S34. For j∈Γ, variable ξ _j is an infeasible solution. Further update κ and

Among them, R is a non-empty subset and

S35. When the update rule falls into a loop or cannot find the corresponding solution, use equation (12) as a candidate update rule to find a feasible solution, that is:
R＝{j:j＝max{j∈Γ}} (12)

S36. In the update process, introduce a buffer factor As infeasible solutions increase, the buffer factor decreases. If the buffer factor is 0, the candidate update rule is used to find a feasible solution and stored in β.

S37: Perform cyclic updates based on S32 to S36 until all solutions to the variables to be obtained are feasible solutions.

The detailed pseudo code of the block pivot rotation method is shown in Figure 3.

The step S4 specifically includes:

S41. Initial solution obtained according to S36 Further construct the loss function, namely:

S42. Use the first-order Taylor series expansion formula to exist near expand, where Represents the first two terms of variable θ ^t ; represents the final estimate. Then the corresponding expression can become:

in,

S43. Combine steps S41 and S42 to obtain the function. The loss function can be further expressed as:

S44, get the expression through S43, for To find the partial derivative, then:

S45. Let the expression obtained in S44 be 0, and the error can be obtained. Then perform the error correction operation to obtain:

In order to verify the effectiveness of the algorithm LCCT-IF provided by this invention in positioning when a shipwreck occurs, a simulation experiment was conducted in Matlab R2021b, and a random walk model was used to simulate the dynamics of the ocean height, so that the positions of all targets and anchor nodes are changing at each moment. , by comparing different information fusion positioning algorithms, such as weighted least squares (WLS), square distance weighted least squares (SRWLS), linear least squares (LLS), and positive semi-definite programming algorithm (SDP), using the root mean square The error is the evaluation criterion, that is, equation (19), and simulation experiments are conducted under different conditions.

Among them, x ^t represents the true position; represents the estimated position; t _max represents the total time, which is set to 1000s in the simulation.

Figure 4 shows the positioning performance of different monitoring area sizes. Affected by offshore wind currents, etc., the falling target will have a dynamic impact as time goes by, so the area of the monitoring area will also show a dynamic change process. Set the side length of the monitoring area as a variable, and other related parameters are: α ^t =3.5, N=8. Figure 4(a) shows the positioning error under different area side lengths. It can be seen from the figure that the positioning errors of WLS, SRWLS and SDP increase as the monitoring area increases. However, LLS and the LCCT-IF algorithm proposed by the present invention have a certain degree of robustness to changes in the monitoring area, and their positioning errors are maintained at a relatively stable level. The LCCT-IF algorithm of the present invention has better positioning accuracy than LLS. As can be seen from Figure 4(b), LCCT-IF can maintain an error within 5m with a probability of 95% in different monitoring areas. Compared with the best-performing LLS among other algorithms (which achieves an error of 95% with the same probability) For 7m) it will be better.

Figure 5 shows the positioning performance of different noises. Since the maritime environment is complex and changeable, its channel conditions are relatively complex. To this end, the positioning accuracy under different noise conditions needs to be verified. The simulation is carried out in a square monitoring area with a side length of 100m. The relevant parameters are specifically set as follows: N=8, ^αt =3.5, It can be seen from Figure 5(a) that the positioning error increases with the increase of noise. The LCCT-IF proposed by the present invention has better performance than other methods, and its positioning error can always be maintained within 3m. In addition, it can be seen from Figure 5(b) that the algorithm proposed by the present invention can achieve 95% probability of errors less than 1.89m, 3.32m, 4.16m and 5.06m under different noise conditions, while other methods achieve the same probability. The errors all exceed LCCT-IF.

Although the content of the present invention has been described in detail through the above preferred embodiments, it should be understood that the above description should not be considered as limiting the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above. Therefore, the protection scope of the present invention should be defined by the appended claims.

Claims (5)

1. An information fusion wireless sensor network positioning method for maritime search and rescue, which is characterized by including the following steps:

   S1. Considering the wave shadowing noise, construct the RSS and TOA ranging models respectively;

   S2. According to the constraints, integrate the multi-source ranging information in the RSS ranging model and the TOA ranging model, and construct a hybrid ranging constrained least squares framework;

   S3. Introduce a buffer factor and use a method to improve block pivot rotation to obtain the initial position of the target. Specifically: divide the variable index to be found into two sets and define a subset of the corresponding variables, calculate its complementary basis solution and judge Whether it is a feasible solution, if not, the set is updated, and the update process introduces a buffer factor; the above judgment and update steps are alternately cyclically alternated, and finally all solutions to the variables to be found are feasible solutions, that is, the initial solution of the target position;

   S4. Error correction is performed on the initial solution described in S3. First, a loss function is constructed, and then linear re-optimization is performed based on the first-order expansion of the Taylor series. Through error correction, an error-corrected solution with a more accurate position is obtained.
2. An information fusion maritime search and rescue wireless sensor network positioning method as claimed in claim 1, characterized in that, in step S1, considering the wave shielding noise, RSS and TOA ranging models are constructed respectively, specifically including:

   S11. If there are N anchor nodes in the network, the position of the i-th anchor node at time t can be expressed as The position of the target to be rescued at time t is expressed as The anchor node can receive signal strength information from the target propagated through radio signals at each moment, forming an RSS ranging model that uses the signal strength information propagated by radio signals to measure distance, that is:
   

   in, Indicates the transmit power value of the target received by the i-th anchor node at time t; d ₀ indicates the reference distance value, usually 1m; represents the target’s transmit power at time t; PL(d ₀ ) represents the loss value of the transmitted signal strength at the relevant reference distance; α ^t represents the path loss factor at time t; Represents the second-order norm; Indicates that the mean of the RSS ranging model is zero and the variance is Wave masking noise that satisfies Gaussian distribution;

   S12. Similarly, the arrival time information from the target is received through the anchor node, and a TOA ranging model is constructed based on the signal propagation time, which is expressed as:
   

   in Indicates that the mean value of the TOA ranging model is zero and the variance is respectively Wave obscuring noise satisfying Gaussian distribution.
3. An information fusion maritime search and rescue wireless sensor network positioning method as claimed in claim 2, characterized in that step S2 fuses the ranging information in the RSS and TOA ranging models, and combines the constraints to construct hybrid ranging constraints. The least squares framework specifically includes:

   S21. According to the expression (1a) obtained in S11, we can obtain:
   

   in,

   S22. Linearly expand the equation (2) obtained in S21 to get:
   

   S23. Consider constraints as well as Combining Equation (1b), the original positioning problem is constructed as an unconstrained least squares framework expression based on hybrid ranging, that is:
   

   S24. Perform square expansion on equation (4) obtained in S23 to get:
   

   S25, order As a variable, where M represents the quantity to be found in the variable, combined with the constraint θ ^t ≥ 0, the original positioning problem can be further constructed as a hybrid ranging constrained least squares (HM-CLS) framework, that is:
   

   in,
   

   Among them, A ^t and B ^t are the matrix parameter values of the hybrid ranging constrained least squares (HM-CLS) framework.
4. An information fusion wireless sensor network positioning method for maritime search and rescue as claimed in claim 3, characterized in that, in step S3, a buffer factor is introduced and an improved block pivot rotation method is used to obtain the initial position of the target, specifically including:

   S31. Divide the index of the variable M to be found in S25 into two sets, namely κ and in In addition, define ξ _κ , Ψ _κ , is a subset of the corresponding variables, and is the submatrix of the corresponding matrix A ^t ;

   S32. Calculate the complementary basis solution according to formula (8) and formula (9) If the constraint ξ _κ ≥ 0 is satisfied and If so, it is a feasible solution, otherwise it is an infeasible solution;
   
   

   S33. When the complementary base solution does not satisfy the constraints, define the set Γ to satisfy:
   

   S34. For j∈Γ, the variable ξ _j is an infeasible solution; further update κ and
   

   Among them, R is a non-empty subset and

   S35. When the update rule falls into a loop or cannot find the corresponding solution, use equation (12) as a candidate update rule to find a feasible solution, that is:
   R＝{j:j＝max{j∈Γ}} (12)

   S36. In the update process, introduce a buffer factor When the infeasible solutions increase, the buffer factor decreases; if the buffer factor is 0, the candidate update rule is used to find feasible solutions and stored in β;

   S37: Perform cyclic updates according to S32 to S36 until all variable solutions to be found are feasible solutions, that is, initial solutions for the target position.
5. An information fusion maritime search and rescue wireless sensor network positioning method as claimed in claim 4, characterized in that, in step S4, based on the first-order expansion of Taylor series, a linear re-optimization method is derived, and in step S36 The obtained initial solution is error corrected to obtain a more accurate solution, including:

   S41. Initial solution obtained according to S36 Further construct the loss function, namely:
   

   S42. Using the first-order Taylor series expansion, exist near expand, where Represents the first two terms of variable θ ^t ; represents the final estimated value; the corresponding expression can become:
   

   in,
   

   S43. Combine steps S41 and S42 to obtain the function. The loss function can be further expressed as:
   

   S44, get the expression through S43, for To find the partial derivative, then:
   

   S45. Let the expression obtained in S44 be 0, and the error can be obtained. Then perform the error correction operation to obtain:
   

   Right now is the error-corrected solution.