CN113507744B - Precision-adjustable distributed wireless network target positioning method - Google Patents

Abstract

The invention discloses a precision-adjustable distributed wireless network target positioning method, which comprises the following implementation steps: step 1: setting an initial target position guess

Step 2: each wireless positioning node k obtains a noisy distance value by collecting signal strength RSSI

And step 3: at the current iteration time i, each wireless positioning node k calculates the error between the node k and the target

And 4, step 4: at the current iteration time i, calculate it relative to

Gradient vector of

Thereby obtaining a target estimation value at the next time

And 5: judgment of

If true, the algorithm ends; step 6: setting temp as temp +1, each wireless positioning node k exchanges with neighbor node set N _ k in its communication range

And 7: performing a weighted consistency operation, step 8: updating the iteration

And step 9: if temp. is detected<Returning to the step 6 if t is not equal, otherwise, executing the step 10; step 10: and updating the iteration times and returning to the step 2. The algorithm in the invention is a distributed wireless network positioning method with adjustable precision and high consistency.

Description

Precision-adjustable distributed wireless network target positioning method

Technical Field

The invention relates to the technical field of wireless sensing technology and wireless sensing network positioning, in particular to a wireless sensing network node positioning method.

Background

In a wireless network environment, target positioning has wide application and important significance. For example, the location of mobile phone signals in wireless cellular networks, and the location of various sensing objects in wireless sensor networks. In these applications, positioning accuracy as a core index is particularly important.

The current target positioning method mainly comprises two technical schemes, one is free of ranging, and the other is based on ranging. The former does not need to measure the distance between an anchor node (a network node for positioning) and a target, and the general positioning precision is not high, but the calculation and communication costs are lower; the latter requires measuring the distance between the anchor node and the target, and has relatively high positioning accuracy and high cost of calculation and communication. With the rapid development of wireless communication and electronic computing technologies, the current target positioning algorithm has higher and higher requirements on positioning accuracy without excessive consideration of cost.

The main method for measuring the distance of the wireless Signal is to adopt a received Signal Strength rssi (received Signal Strength indicator) technology, which is greatly influenced by environmental noise. The higher the measurement accuracy when the signal to noise ratio is high. Under the extreme condition of no noise, the target can be accurately positioned only by using three anchor nodes and simple trilateral positioning. In contrast, in a high noise (low signal-to-noise ratio) environment, more anchor nodes are needed to form a wireless network to resist the influence of noise, and the target position is estimated cooperatively.

Range-based targeting can be modeled as a typical nonlinear Least mean square (NLLS) (nonlinear Least squares) convex problem with a globally optimal solution, and thus can be solved using unconstrained optimization methods. Generally, either first order gradient descent or second order Newton-like methods can be developed. However, the gradient descent method has too slow a convergence rate, and the second-order Newton method is computationally too extensive. Recently, the inventor develops a target positioning algorithm based on a Gauss-Newton method in a high-noise industrial environment, the algorithm exchanges estimation information with neighbor nodes, a time iteration optimization method based on local communication is adopted, and target estimation information from the neighbor nodes at the current moment is integrated in the local calculation process, so that higher estimation precision is obtained. The algorithm mainly comprises two steps of consistency communication and iterative computation, wherein in the consistency communication process, any anchor node k receives target estimation of a neighbor anchor node j at a time i

And transmits its own target estimation value

Giving a node j and weighting and summing; and in the iterative calculation process, the Gauss-Newton of the depth fusion is implemented to descend.

The generated Diffusion Gauss-Newton (DGN) algorithm is simple to implement, and obtains better comprehensive performance in the similar algorithms, but the following defects still exist:

(1) the positioning accuracy still needs to be improved;

(2) the consistency of each anchor node to the estimation is not high, namely, all the local estimation along with i → ∞cannotbe well satisfied

All converge to the exact neighborhood of the same value;

(3) the precision is not controllable, namely the positioning algorithm cannot flexibly adjust the positioning precision according to the application requirement.

Disclosure of Invention

To overcome the above-mentioned disadvantages of the DGN algorithm, we have found that this problem can be alleviated to some extent by increasing the number of coherent communication rounds, i.e. implementing multiple rounds (1 to t, t being specified by the user) of coherent communication during each iteration, the resulting algorithm can be referred to as DGN^t. When t is 1, DGN^tI.e. the original DGN algorithm. When t is>1, the obtained effect is that the consistency is improved, and the local node can absorb external information more sufficiently, so that the positioning precision in each iteration process can be further improved. Meanwhile, a balance between the estimation accuracy and the communication cost can be dynamically obtained by setting different values of t, that is, the higher the accuracy requirement is, the higher the communication cost (mainly considering the energy consumption generated by signal transmission) is, and the cost can be borne by the user. Therefore, the algorithm in the invention is a distributed wireless network positioning method with adjustable precision and high consistency.

DGN^tAnother significant difference from DGN is the different order of execution of the two steps of coherent communication and iterative computation. The DGN executes consistency step and then executes iterative computation; in the present invention, however, DGN^tIterative computation is first performed and then consistent communication is conducted. The adjustment brings the advantage that the updated data calculated at the current moment is exchanged with the neighbor in the communication process, so that the data timeliness is better.

Precision-adjustable distributed wireless network target positioning method named DGN^tThe algorithm comprises the following implementation steps:

step 1: setting an initial target position guess

The initial round value t and a threshold epsilon, the temporary variable temp 0 and the step parameter alpha e (0, 1)]；

Step 2: each wireless location node k (its own three-dimensional coordinate y)_kFixed and known) obtains noisy distance values to unknown targets (whose true coordinates are assumed to be x) by collecting signal strength RSSI

And step 3: at the current iteration time i, each wireless positioning node k calculates the error between the node k and the target according to the following formula

And 4, step 4: at the current iteration time i, each wireless positioning node k is according to the current error

Calculate its relative to

Gradient vector of

And the following Gauss-Newton stack was performedUpdating generation

Thereby obtaining a target estimation value at the next time

And 5: judgment of

If true, the algorithm ends; otherwise, executing step 6;

and 6: setting temp as temp +1, each wireless positioning node k exchanges with neighbor node set N _ k in its communication range

And 7: the following weighted coherency operation is performed

Wherein the weight parameter w_k,lThe conditions need to be satisfied: when in use

Time w _k,l0 and

w_kl≥0。

and 8: updating the iteration estimation value

And step 9: if temp is less than or equal to t, returning to step 6, otherwise executing step 10;

step 10: and updating i to i +1, and returning to the step 2.

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

through simulation experiments, compared with the DGN algorithm (namely DGN)¹) DGN in the present invention^tThe algorithm is obviously improved in three aspects of positioning precision, consistency estimation and precision adjustment. The simulation parameters are set as follows: the k is 20 positioning nodes to simulate anchor nodes, and the random deployment is 100 × 100m in the region³The initially guessed positioning coordinates are unified into

Randomly selecting the real coordinate of the target; for continued observation of the experimental results, the threshold epsilon is not specified in this experiment, but a fixed number of iterations 2000 is used to terminate the iteration; the step parameter alpha is 0.01; simulating to generate Gaussian environmental noise with the average value of 0 and the standard deviation of 0.1; finally, the round number t of the coherent communication takes values of 1, 5 and 10 respectively for comparison.

As can be seen from the experimental results, firstly, from the aspect of positioning accuracy (fig. 3 is a sampling graph of 2000 iteration results, and fig. 4 is a complete graph displayed in a similarity manner in fig. 3), as the algorithm is iterated, the positioning errors obtained by the three algorithms are reduced continuously; second from steady state accuracy, DGN¹⁰Is superior to DGN⁵And DGN⁵Is superior to DGN¹But DGN¹⁰Specific DGN⁵The advantages of (2) are not obvious, and show that the more the number of rounds of consistent communication is, the better the number is, certain compromise needs to be adopted, so that the controllability of the algorithm between the positioning accuracy and the communication cost is also reflected; finally, FIG. 5 shows the standard deviation of the positioning error between all nodes for the three algorithms over 2000 iterations, from which it can be seen that the DGN¹Is the largest, and the DGN proposed in the present invention⁵And DGN¹⁰DGN as represented^tThe algorithm has a smaller standard deviation, which means that the consistency purpose of the algorithm is better realized.

Drawings

FIG. 1 is a DGN of the present invention^tAn algorithm flow chart;

figure 2 is a schematic diagram of a handset signal location for a wireless cellular network of the present invention;

FIG. 3 is a DGN of the present invention^tPositioning accuracy sampling chart of algorithm, displaying DGN^tThe algorithm applies t values of 10, 5 and 1 (i.e., DGN)¹⁰，DGN⁵And DGN¹) Along with a positioning precision graph of algorithm iterative convergence, the horizontal coordinate is a time iteration step, and the vertical coordinate is positioning precision;

FIG. 4 is a DGN of the present invention^tThe similarity graph of the algorithm positioning accuracy is used for more clearly displaying the difference of the three algorithms in the graph 3 so as to facilitate performance comparison;

FIG. 5 is a DGN of the present invention^tThe positioning standard deviation sampling graph of the algorithm, namely the difference between the positioning accuracy of all nodes along with the operation of the algorithm, describes the index of the consistency of the algorithm.

Detailed Description

The invention is further described with reference to the accompanying drawings in which:

referring to fig. 1 and fig. 2, a specific implementation process of the algorithm is shown by taking the positioning of a mobile phone signal in a wireless cellular network as an example. The reason for taking the wireless cellular network as an example is that, with the popularization of 5G base station signals in the future, the density of 5G base stations will be greater due to the faster frequency of the 5G signal compared with the frequency of the 4G signal, and the wireless cellular network is also more suitable for the co-location among the base stations to resist the ranging noise. Of course, the present invention is suitable for all network environments that are co-located in a wireless signal ranging manner, such as the location of a sensor network to a target in various environments (e.g., an industrial environment, a logistics environment, etc.), the location of a target in a local area network multi-WIFI environment, and so on.

Step 1: setting the same initial conditions for all wireless base stations

t, ε and α ∈ (0, 1)]；

Step 2: a plurality of wireless base stations serve as anchor nodes to collect the RSSI (received signal strength indicator) of the target mobile phone signal strength, and noisy distance values with errors are obtained;

and step 3: at iteration time i, each wireless base station obtains a target cost function through calculation

And 4, step 4: at iteration time i, each radio base station is based on

Implementing Gauss-Newton updating to obtain a new position estimation value;

and 5: each base station calculates the Euclidean distance between the position estimation of the two moments before and after and judges the magnitude of the Euclidean distance and a threshold value epsilon; if the value is less than epsilon, the algorithm is ended, otherwise, the next step is executed;

step 6: each wireless base station sends an estimation value to an adjacent base station, and simultaneously, the estimation values from all the adjacent base stations are also obtained, so that the exchange of the estimation values is realized;

and 7: each wireless base station performs a round of consistency weighted average to obtain

And 8: updating

And step 9: judging temp is less than or equal to t, if true, returning to the step 6, and continuing to implement a new round of consistency weighted average; otherwise, executing step 10;

step 10: and updating i to i +1, and returning to the step 2.

Claims (1)

1. A precision-adjustable distributed wireless network target positioning method is characterized by comprising a DGN^tNamely, the Diffusion Gauss-Newton algorithm of t-round communication, the implementation steps are as follows:

step 1: setting an initial target position guess

Initial round value t and threshold epsilon, temporary variable temp 0 and step size parameter alpha e (0, 1)]；

Step 2: each wireless positioning node k is known with a three-dimensional coordinate y fixed by itself_kAcquiring a noisy distance value between the time i and an unknown target by collecting a Signal Strength RSSI (Received Signal Strength Indicator)

The true coordinates of the target are assumed to be x;

and step 3: the estimation value of each wireless positioning node k to the target at the current iteration moment i is

And calculating the error between node k and the target according to the following formula

And 4, step 4: at the current iteration time i, each wireless positioning node k is according to the current error

Calculate its relative to

Gradient vector of

And the following Gauss-Newton iterative update is performed

Thereby obtaining the target estimated value of the next iteration moment

And 5: judgment of

If true, the algorithm ends; otherwise, executing step 6;

step 6: setting temp as temp +1, each wireless positioning node k exchanges with neighbor node set N _ k in its communication range

And 7: the following weighted coherency operation is performed

Wherein the weight parameter w_k,lThe conditions need to be satisfied: when in use

Time w_k,l0 and

and w_k,l≥0；

And 8: updating the iterative estimation

And step 9: if temp is less than or equal to t, returning to step 6, otherwise executing step 10;

step 10: and updating i to i +1, and returning to the step 2.

Images