Target positioning method based on virtual anchor points under narrowband signal condition

Abstract

According to the target positioning method based on the virtual anchor point under the narrowband signal condition, under the condition of a single base station and a single antenna, the initial position of the target is not needed, and the target is positioned only through TDOA. And because the single-dimensional parameter estimation error of the narrowband signal is large and the multipath capability resolution is not strong, the positioning error of a single point is large. Therefore, the method combines historical information of target motion, adopts the thought of traversal search, and abandons conversion from single-point positioning to track positioning. According to the relative displacement and observation data of each sampling point of a target in a period of time, the positions of K points are reversely deduced on one particle, and then the same trajectory is generated. Meanwhile, a TDOA set of each point on the generated trajectory is calculated through VA, finally, the sum of the TDOA of all sampling points on the whole trajectory and the matching loss value of observation data on the real trajectory is calculated with the help of a Hungary algorithm, and the point corresponding to the trajectory with the minimum loss value is the position of the target. In addition, the PSO algorithm is used, so that the time for traversing and searching on the map is greatly shortened, and the real-time performance of the system is improved.

Description

Technical Field

[0001] The invention belongs to indoor positioning technology, which is a target positioning method based on virtual anchor points under narrowband signal conditions. Background Art

[0002] In the current outdoor positioning, satellite positioning systems such as BeiDou Navigation Satellite System (BDS) and Global Positioning System (GPS) have matured and can provide high-precision positioning in outdoor environments. However, in indoor environments or areas where signals are blocked (such as underground parking lots or shopping malls), the signal attenuation is large, the error is large, and the positioning accuracy cannot meet the actual positioning needs. Therefore, indoor positioning technology has become a key solution to fill this gap and has become a hot research topic. In previous studies, most indoor positioning algorithms obtain wireless parameters such as time of flight (TOF), angle of arrival (AOA) and angle of departure (AOD) through multiple base stations (BS), and then process these parameters to achieve target positioning. However, in actual application scenarios, implementing a high-precision multi-base station positioning system usually significantly increases hardware requirements and costs, especially in large or complex environments. In addition, base stations in a multi-base station system usually need to be precisely synchronized, and the clock asynchrony between base stations may cause significant errors in positioning calculations. Therefore, positioning technology based on a single base station is particularly necessary.

[0003] Multipath-assisted technology provides an effective solution for single-station positioning. Multipath signals can be divided into single reflection paths, multiple reflection paths, scattering, etc. Among them, the single reflection path is the most special one. It is obtained by the mirror reflection of the signal on the reflector. It is a special case that satisfies the angle of incidence equals the angle of reflection. Therefore, the base station can be mirror-symmetric about the reflector to obtain a virtual anchor (VA). VA acts as a mirror of the base station. At this time, the single reflection path can be regarded as the signal emitted by VA, thereby providing a strong condition for positioning. In actual positioning, only the line of sight (LOS) path and the single reflection path are useful information. Their actual receiving power is relatively large and can be received at the receiving end, while the remaining multipath power is small and is regarded as clutter. In reality, due to the irregularity of the terminal and base station antenna arrays, classic angle estimation algorithms, such as MUSIC and ESPRIT, are difficult to use in practical applications, which makes it difficult to use the angle information of multipath. TOF has no special requirements for the antenna array, and a single antenna can complete TOF measurement. However, in actual tests, wireless signals will have propagation delays, and the time difference of arrival (TDOA) between multipath signals and LOS paths is accurate. Therefore, under the condition of a single base station and a single antenna, it is particularly important to use TDOA positioning.

[0004] The present invention proposes a virtual anchor point positioning method based on narrowband signals, which realizes the positioning of the target by only using TDOA under the condition of a single base station and a single antenna. The traditional positioning method based on VA realizes the positioning of the target by using ultra-wideband (Ultra-Wideband, UWB) signals. UWB can provide high-precision parameter estimation, and uses high-precision TOF and angle information to complete the positioning of the target. The positioning method based on narrowband signals proposed by the present invention realizes the positioning of the target by using multipath TDOA through the use of VA in the case of inaccurate parameter estimation. Since the requirements for TDOA in single-point positioning are high, and the actual TDOA accuracy cannot meet the requirements, the present invention combines the historical information of the target movement, adopts the idea of ​​traversal search, and abandons the transformation of single-point positioning into trajectory positioning. Through the relative displacement and observation data of each sampling point of the target within a period of time, the positions of the remaining points can be inferred from the position of one point to form the same trajectory. At the same time, the TDOA set of each point on the trajectory generated by VA calculation is used, and finally the sum of the loss values ​​of the entire trajectory matching is calculated with the help of the Hungarian algorithm. The point with the smallest loss value is the position of the target. However, the complexity of global traversal search increases exponentially with the improvement of environment size and traversal accuracy. Therefore, the present invention adopts particle swarm optimization (PSO) algorithm, which can greatly reduce the search time, reduce the computational complexity, and improve the real-time performance of the system. In addition, the present invention avoids using only the loss value in data association, avoids the mismatch problem caused by association failure, and completes the positioning of the target without the initial position of the target. Summary of the invention

[0005] The purpose of the present invention is to provide a target positioning method based on virtual anchor points under narrowband signal conditions, which can effectively utilize multipath TDOA and historical information of the target to complete the positioning of the target.

[0006] The virtual anchor point positioning method based on narrowband signals of the present invention comprises the following steps:

[0007] Step 1: Record the relative displacement and observation value of the target K-1 times before;

[0008] Step 2: Set the number of particles N in the particle swarm PSO , maximum number of iterations J, learning factors c1 and c2, maximum value of inertia weight ω max and the minimum value of the inertia weight ω min and the scope of the search space;

[0009] Step 3: Randomly initialize the position and velocity of the particle in the search space and calculate the fitness value of the particle;

[0010] Step 4: Initialize the individual optimal solution of each particle and the group optimal solution of the particle swarm;

[0011] Step 5: Update the speed and position of each particle and calculate the fitness value of the particle based on the observation values ​​at the current moment and the previous K-1 moments;

[0012] Step 6: Update the individual optimal solution and group optimal solution of each particle;

[0013] Step 7: Determine whether the maximum number of iterations has been reached. If so, the optimal solution for the group at this time is the target position. If not, return to step 5.

[0014] Beneficial Effects

[0015] The present invention is a target positioning method based on virtual anchor points under narrowband signal conditions, which has the following advantages:

[0016] 1. Positioning can be done without the target’s initial position;

[0017] 2. The target can be positioned with a single base station and a single antenna;

[0018] 3. Only multipath TDOA is used to avoid the delay error problem of TOF;

[0019] 4. The target can be positioned even when the parameter estimation error is large. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of the present invention;

[0021] Figure 2 Schematic diagram of the positioning method. Specific implementation plan

[0022] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0023] Figure 1 It is a flow chart of the present invention, and the specific steps are as follows:

[0024] Step 1: Model the motion of the target. The state of the target can be expressed as where x t and t are the two-dimensional coordinates of the target at time t, and are the speeds of the target in the x and y directions at time t, respectively. The state transfer equation of the target is:

[0025]

[0026] When the target starts to move, it is not positioned, and the relative displacement and observation value of the target at the previous t-1 moments are recorded according to the state transfer equation.

[0027] Step 2: Set the number of particles N in the particle swarm PSO , maximum number of iterations J, learning factors c1 and c2, maximum value of inertia weight ω max and the minimum value of the inertia weight ω min and search space range, where c1 is used to control the movement of particles to the individual optimum, and c2 is used to control the movement of particles to the group optimum. In general, in order to avoid excessive particle movement or falling into the local optimum, c1+c2≤4 is controlled to ensure that the particle swarm has appropriate exploration and development capabilities; for inertia weight, a larger inertia weight emphasizes global search capabilities, making it easier for particles to explore unknown areas, and the convergence speed may be slower, while a smaller inertia weight emphasizes local search capabilities, and its convergence speed is faster but it is easy to fall into the local optimum. Therefore, a linear decreasing weight strategy can be adopted, gradually reducing from a larger weight value to a smaller weight value, which can make the particle swarm more exploratory in the early stage and pay more attention to local optimization in the later stage.

[0028] Step 3: Randomly generate N PSO particles, and calculates their fitness values ​​based on the relative displacement of the generated particle positions and the target. Figure 2 As shown in the figure, under the randomly generated particle position, the relative displacement of the target at the previous K-1 moments is recorded, and the positions of K-1 points are inferred. The fitness values ​​of the positions of K points are calculated and added using the Hungarian algorithm to obtain the fitness value of the particle. The specific steps are as follows:

[0029] 3a. Calculate the positions of K-1 points based on the recorded relative displacements;

[0030] 3b. Calculate the TOF from the K points to all VAs and base stations, and then subtract the TOF to the base station from the TOF to the VA to obtain the TDOA set at that point.

[0031] 3c. Subtract the TDOA in the actual observation set from each element in the TDOA set calculated in the previous step to obtain an N×M-dimensional cost matrix, where N is the number of elements in the calculated TDOA set and M is the number of elements in the actual observation set.

[0032] 3d. Determine whether the cost matrices N and M are equal. If they are not equal, fill the matrix with 0s according to the maximum value of the two to make the matrix with the same number of rows and columns;

[0033] 3e. Subtract the minimum value of each row of the cost matrix so that at least one element in each row is 0, and subtract the minimum value of each column of the cost matrix so that at least one element in each column is 0;

[0034] 3f. Find independent 0s in the matrix and draw lines (horizontally or vertically) to cover all 0s as much as possible. If the number of lines is equal to the dimension of the matrix, the optimal solution is found and the matching is completed. Otherwise, proceed to the next step.

[0035] 3g. Find the minimum element not covered by the line, subtract the minimum value from the uncovered element, add the minimum value to the cross-covered element, generate a new cost matrix, and return to the previous step.

[0036] After executing the above steps, the matching result can be obtained. According to the matching result of each point, the TDOA difference of each point is calculated in turn and added up to get the fitness value of the point. Then, the fitness value of K points is added to get the fitness value of the particle.

[0037] Step 4: According to N PSO The fitness value of each particle is initialized to the individual optimal solution of each particle and the optimal solution of the particle swarm g best The individual optimal solution is the coordinate with the minimum fitness, and the group optimal solution is N PSO The coordinates of the particle with the smallest fitness among the particles.

[0038] Step 5. First, update the inertia weight according to the current number of iterations:

[0039]

[0040] Where j represents the current iteration number, and then updates the particle speed:

[0041]

[0042] Where r1 and r2 are random numbers distributed between [0,1]. is the position of the particle after the jth iteration, and finally the position of the particle is updated:

[0043]

[0044] After obtaining the updated particle position, the fitness value of each particle is calculated according to the particle fitness calculation method in step 3.

[0045] Step 6: According to the updated fitness value of each particle, update the individual optimal solution of each particle and the group optimal solution of the particle swarm again.

[0046] Step 7: Determine whether the number of iterations reaches the maximum value. If it reaches the maximum value, output the optimal solution of the particle swarm as the positioning result. If it does not reach the maximum value, return to step 5.

Claims

1. A target positioning method based on virtual anchor points under narrowband signal conditions, characterized in that: a) The Hungarian algorithm is used to calculate the fitness function value of the particle swarm optimization algorithm (PSO); b) Use PSO algorithm to search for the location of the target in the map; c) When the error of narrowband signal parameter estimation is large, the historical data is used to locate the target; d) The target can be located without the initial position of the target.

2. According to claim 1, the fitness function value of PSO is calculated using the Hungarian algorithm, characterized in that The loss value of the PSO search is obtained. In the PSO algorithm, the fitness function is its core. It mainly measures the quality of the current particle position, provides the algorithm with optimization goals and directions, and enables the particle swarm to gradually evolve towards the global optimal solution. First, it is necessary to construct a cost matrix. In the case of a single base station, the observations received by the target at each moment include the line of sight (LOS) path, single reflection path, multiple reflection path, scattering and clutter, etc. However, due to the delay of the wireless signal in the actual test, the test signal flight time (TOF) has errors, which further leads to inaccurate positioning. The time difference of arrival (TDOA) between the multipath signal of the wireless signal and the LOS path is accurate, so the actual observation is composed of TDOA. In the observation, the single reflection path is caused by the mirror reflection of the signal on the reflector, which corresponds to the virtual anchor point (VA), and the position of VA can be calculated by the base station about the mirror symmetry of the reflector. Therefore, according to the position of the particle in PSO, the TOF between the particle and the base station (BS) and VA can be calculated, and then TDOA can be obtained, so the cost matrix can be constructed. Let each element in the calculated TDOA set be subtracted from each element in the actual observation set in turn, so that an N×M-dimensional cost matrix can be obtained, where N is the number of elements in the calculated TDOA set, and M is the number of elements in the actual observation set. The loss value calculation steps are as follows: a) Determine whether the cost matrices N and M are equal. If not, fill the matrix with 0s according to the maximum value of the two to make the matrix with the same number of rows and columns; b) Subtract the minimum value of each row of the cost matrix so that at least one element in each row is 0 and subtract the minimum value of each column of the cost matrix so that at least one element in each column is 0; c) Find independent 0s in the matrix and draw lines (horizontally or vertically) to cover all 0s as much as possible. If the number of lines is equal to the dimension of the matrix, the optimal solution is found and the matching is completed. Otherwise, proceed to the next step. d) Find the minimum element not covered by the line, subtract the minimum value from the uncovered element, add the minimum value to the cross-covered element, generate a new cost matrix, and return to the previous step. After the matching is completed, according to the matching results, the differences between the matched elements in the two sets are calculated in turn and summed up, which is the final loss value of the Hungarian algorithm and also the fitness function value.

3. According to claim 2, the location of the target in the map is searched by the PSO algorithm, characterized in that Reduce computational complexity and improve real-time positioning. The computational complexity of searching for the target's location by traversing the entire map with a certain accuracy increases exponentially with the size of the map and the accuracy of the search, and cannot meet the real-time requirements of the system. The PSO algorithm can solve this problem by simulating information sharing and collaboration between individuals in a group to achieve global optimization of the problem. PSO generates random particles on the map. The position of the particles represents the solution to the problem. The speed of the particles controls their movement in the solution space, and adjusts their position and speed based on the particles' own historical experience and the optimal solutions of other particles. The ultimate goal of all particles is to find the global optimal solution of the fitness function in continuous iterations, and all particles converge to a point or area in the solution space. The algorithm steps are as follows: a) Set the number of particles N in the particle swarm PSO , maximum number of iterations J, learning factors c1 and c2, maximum value of inertia weight ω max and the minimum value of the inertia weight ω min and the scope of the search space; b) Randomly initialize the position and velocity of particles in the search space; c) calculating the fitness value of the current position of the particle according to the algorithm in claim 1; d) Update the individual optimal solution for each particle And update the global optimal solution g of the entire particle group best , where i represents the i-th particle; e) Update the formula based on inertia weight: Calculate the inertia weight of the current iteration number, where j represents the current iteration number; f) Update the formula according to the speed: Update the particle speed of the current iteration, where r1 and r2 are random numbers distributed between [0,1]. is the position of the particle after the jth iteration, c1 is used to control the particle's movement force toward the individual's optimal state, and c2 is used to control the particle's movement force toward the group's optimal state; g) Finally, update the formula according to the position: Update the position of the particle, calculate the fitness of the particle according to the fitness function, and update the optimal solution of the individual and the global optimal solution g of the group best ; h) Repeat steps e) to g) for each particle in sequence, and finally determine whether the number of iterations reaches the maximum value. If it reaches the maximum value, the algorithm ends; if it does not, return to step e). The final global optimal solution g best , which is the searched target location.

4. According to claim 3, when the error of narrowband signal parameter estimation is large, the historical data is used to locate the target, characterized in that When the data positioning error of a single point is large or even impossible to locate, the target can be located by combining the observations of the previous time period. Under the condition of narrowband signal, the single-dimensional parameter estimation of the signal faces the problem of large error or multipath information loss, which will lead to large error of TDOA and fail to meet the accuracy required for positioning. Therefore, the target cannot be located when it starts to move. It is necessary to use the state transfer equation of the target after the target has moved for a period of time and record the relative displacement of the target at each sampling moment for positioning. Therefore, the state of the target is modeled. where x t and t are the two-dimensional coordinates of the target at time t, and are the speeds of the target in the x and y directions at time t, respectively. The state transfer equation of the target is: At this time, according to the recorded observations of the previous K-1 moments and the relative displacement of the target at the previous K-1 moments, the fitness of each particle can be calculated in the PSO search based on the current particle position and the recorded relative displacement of the target at the previous K-1 moments. The relative displacement is applied to the particle, so the positions of K points can be calculated on one particle to generate the same trajectory, and then the sum of the loss values ​​of the K points is calculated. At this time, the PSO search changes from point search to trajectory search, which can avoid the problem of excessive parameter estimation error or insufficient information for a single point, and then search for the position of the target. After calculating the position of the target at the current moment, the target position at the next moment must be calculated. The target is located by sliding windows again with the help of the observations at the next moment and the observations at the previous K-1 moments.

5. According to claim 4, the positioning of the target can be completed without the initial position of the target, characterized in that The target is positioned during its movement, and the initial position of the target is deduced based on the recorded relative displacement. It can be seen from claim 3 that the positioning of the target at the tth moment can be completed through the observations of the target at the previous K-1 moments. At this time, based on the recorded relative displacement of the target at each moment, the position of the target at the previous K-1 moments can be inferred, and the initial position of the target is thus found.

Images