Indoor positioning method for adaptive data processing

Abstract

The invention discloses an indoor positioning method for adaptive data processing, and relates to the field of data processing. Compared with a traditional classical kernel adaptive filtering method, the method adopts a weight updating strategy, and has higher and more stable filtering precision. According to the invention, the q-series fitting is introduced into the kernel adaptive filter for the first time to construct the positioning system; experiments show that compared with the current most advanced positioning algorithm, the algorithm provided by the invention can obtain better positioning precision. The kernel adaptive filter provided by the invention and the classical kernel adaptive filtering algorithm are applied; compared with an indoor positioning algorithm based on machine learning, the classical indoor positioning method has the highest filtering precision and higher robustness. Compared with a classic kernel adaptive filtering algorithm, a classic indoor positioning method and an indoor positioning algorithm based on machine learning, the high-precision positioning adaptive data processing method has remarkable advantages in the aspect of positioning prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the field of data processing. Background Art

[0002] The RSS fingerprint method is a typical survey-based method that establishes a database of signal features at various indoor locations and matches real-time data with the database to estimate the location. It is often used in Wi-Fi and BLE mapping. Therefore, survey-based methods such as RSS fingerprinting generally have better positioning effects than model-based methods and have become the mainstream method in the field of indoor positioning. However, since the fingerprint method only establishes an RSS dataset of a limited number of discrete location points in the room, an estimation method is required to learn from the dataset and predict the unknown location when the RSS value of the unknown location is given. In recent years, due to the rapid development of deep learning and machine learning, the so-called probabilistic fingerprint (P-FT) method has been formed. There are some typical P-FP methods in the field of indoor positioning, such as S.Gui, Y.Tang, X.Li, D.Bi, L.Peng, Y.Xie, A fractional-derivative kernel learning method for indoor position prediction, ExpertSystems with Applications 252(2024)124133. Among these methods, kernel adaptive filtering may be the most promising one. Compared with traditional machine learning methods, kernel adaptive filtering methods have certain accuracy and robustness. Compared to classic deep learning methods, KAFs-based methods have a lighter network structure, which is crucial for indoor online positioning and reduces hardware burden. A technology closest to ours is indoor positioning based on the K-nearest neighbor (KNN) algorithm. This method applies the K-nearest neighbor algorithm in machine learning to the RSS fingerprint indoor positioning problem, but this technology suffers from low positioning accuracy and poor robustness. Summary of the Invention

[0003] The technical problem to be solved by the invention is that the present invention proposes an adaptive data processing method for high-precision and high-robust indoor positioning. The method is based on a kernel adaptive filter (qKRMC), which can provide high-precision, strong robustness and adaptive data processing capabilities under limited computing resources.

[0004] The technical solution of the present invention is: an indoor positioning method for adaptive data processing, the method comprising:

[0005] Step 1: Offline learning phase:

[0006] The actual position of the movable sensor in the target area is known. i The RSSI value measured at the reference point forms the input RSSI data u i ;

[0007] Step 2: Use kernel adaptive filter based on input RSSI data u i Get the estimated position y i , by estimating the position y i With the actual position d i Comparison to obtain the position error e i , and then modify the internal weights of the kernel adaptive filter to achieve the training data sequence {u i ,d i}'s learning;

[0008] Step 3: Online application stage:

[0009] Actual location d i Unknown input RSSI data u i is passed into the fully trained kernel adaptive filter, which gives the estimated position y i , the estimated position y i As the input RSSI data u i The position estimation result is directly output.

[0010] Furthermore, the specific method of the kernel adaptive filter is:

[0011] Step 1: Define a sequence sample set {u(i), d(i)}, where u(i) represents the input RSSI data at time i to be filtered, and d(i) represents the corresponding actual position; define the kernel function k σ (·), kernel function k σ (·) can map the input RSSI data u(i) into a high-dimensional feature space In this paper, we use data processing in high-dimensional space to learn the nonlinear relationship of the input RSSI data u(i); define The high-dimensional input RSSI data u(i) is mapped to the Hilbert space H based on the kernel function; the high-dimensional input RSSI data The relationship with the low-dimensional input RSSI data u(i) is:

[0012]

[0013] Step 2: Define the loss function L of the kernel adaptive filter as:

[0014]

[0015] Among them, ω represents the weight of the kernel adaptive filter, G θ (·) is the Gaussian kernel function, is the position error at time i; λ is the regularization parameter, β is the forgetting factor, λ‖ω j ‖ 2 is the regularization term, q represents the value of the series, ω i represents the updated weight of the kernel adaptive filter at time i, Represents the high-dimensional input RSSI data u(i) mapped to the Hilbert space H based on the kernel function;

[0016] Step 3: Define the updated weights ω of the kernel adaptive filter i :

[0017]

[0018] Among them, Φ i The expansion of the input RSSI data u(i) High-dimensional input RSSI data obtained by mapping in Hilbert space H; ω i To update the weight;

[0019] Step 4: According to the kernel function shown in formula (1), the weights of the modified kernel adaptive filter are continuously updated according to the input RSSI data u(i) and the actual position d(i) to achieve offline learning.

[0020] Furthermore, in the online application phase in step 3, based on the input RSSI data u(i) whose actual position d(i) is unknown, the estimated position y(i) is obtained by the following relationship:

[0021]

[0022] Subsequently, y(i) is directly output as the position estimate of the input RSSI data of the unknown position by the positioning system to realize the positioning function.

[0023] Furthermore, the kernel adaptive filter is a q-order kernel adaptive filter.

[0024] Compared to traditional kernel adaptive filtering, this method uses a weight update strategy, resulting in higher and more robust filtering accuracy. This is the first time that q-series fitting has been incorporated into kernel adaptive filtering to construct a positioning system. Experiments have shown that this algorithm achieves better positioning accuracy than the most advanced positioning algorithms available. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Flowchart of the adaptive data processing method for high-precision positioning.

[0026] Figure 2This is the principle diagram of the kernel adaptive filtering algorithm based on q series fitting.

[0027] Figure 3 For indoor positioning experimental classroom scene.

[0028] Figure 4 Schematic diagram of classroom indoor positioning training data.

[0029] Figure 5 Schematic diagram of laboratory indoor positioning test data.

[0030] Figure 6 Schematic diagram of classroom indoor positioning training data.

[0031] Figure 7 Schematic diagram of classroom indoor positioning test data.

[0032] Figure 8 This is the prediction performance diagram of each algorithm for classroom indoor positioning X coordinate.

[0033] Figure 9 This is the prediction performance diagram of each algorithm for classroom indoor positioning Y coordinate.

[0034] Figure 10 This is the prediction performance diagram of each algorithm for laboratory indoor positioning X coordinate.

[0035] Figure 11 This is the predicted performance diagram of each algorithm for laboratory indoor positioning Y coordinate. DETAILED DESCRIPTION

[0036] The present invention verifies the superiority of the above-mentioned adaptive data processing method for high-precision positioning through indoor positioning experiments, and uses two real-scene indoor positioning data sets to demonstrate and verify the performance of the present invention. The first real scene is a classroom with a lot of interference, such as Figure 3 As shown in Figure 2; the second real scenario is in a laboratory with moderate interference.

[0037] The specific implementation of the high-precision positioning adaptive data processing method proposed in this invention in this experiment is as follows:

[0038] Experimental process and Figure 1 The process of adaptive data processing method for high-precision positioning is shown Figure 1 First, place the signal receivers in the following Figure 4 and Figure 5 At the training data reference point shown in , the RSSI value of each position is recorded to generate input RSSI data, and then the training data sets corresponding to the two experiments are generated respectively. Subsequently, in order to test the performance of the proposed data processing method, the signal receivers are placed as follows Figure 6 and Figure 7At the randomly generated test points shown, the RSSI value at each location is recorded to generate input RSSI data and test datasets corresponding to the two experiments. The training and test datasets contain the data sequence {u(i), d(i)}. The input RSSI data u(i) and the corresponding actual location d(i) are both known in both datasets.

[0039] During the experiment, the kernel adaptive filter of the present invention was first fed with both the input RSSI data u(i) and the corresponding actual location d(i) from the training dataset for filter learning. Subsequently, only the input RSSI data u(i) from the test dataset was fed into the kernel adaptive filter of the present invention to obtain the corresponding estimated location y(i), thereby achieving online application functionality. Finally, the mean squared error (MSE) was used to measure the difference between the actual location d(i) of the test dataset and the estimated location y(i) obtained during the test, reflecting the accuracy of the positioning data processing.

[0040] The experiments were conducted in the two representative indoor positioning experimental scenarios described above to compare the classical kernel adaptive filtering algorithms (KRLS, KLMP, KLMS, KRLP, KRMC), classical indoor positioning algorithms (K nearest neighbor, naive Bayes, trilateration) and machine learning-based algorithms (transfer learning).

[0041] The performance comparison results of the adaptive data processing method for high-precision positioning proposed in this invention and other kernel adaptive filtering algorithms are as follows: Figure 8-11 The results show that the proposed method achieves the highest steady-state filtering accuracy in representative scenarios with varying interference. Compared to the classical kernel adaptive filtering method and the Student Kernel Adaptive Filter (StKAF) method based on integer-order calculus, the proposed method achieves higher filtering accuracy and is more robust.

[0042] The indoor positioning prediction errors and variances of the kernel adaptive filter of the present invention and the comparison algorithms are shown in Tables 1 and 2. The results show that the kernel adaptive filter of the present invention, which applies the proposed method, has the highest filtering accuracy and higher robustness compared with the classic kernel adaptive filtering algorithm, the classic indoor positioning method, and the indoor positioning algorithm based on machine learning.

[0043] In summary, compared with the classical kernel adaptive filtering algorithm, the classical indoor positioning method and the indoor positioning algorithm based on machine learning, the invented adaptive data processing method for high-precision positioning has significant advantages in positioning prediction accuracy.

[0044] Table 1 Prediction error and variance of each algorithm for classroom indoor positioning

[0045] algorithm Prediction error (m) <![CDATA[Variance (m 2 )]]> Kernel adaptive filter of the present invention 1.8386 1.0601 KRLS 1.9054 1.1445 KLMP 2.2527 0.8605 KLMS 2.3440 0.9109 KRLP 1.8644 0.8550 KRMC 1.9036 1.0229 K-nearest neighbor algorithm 1.9446 1.1440 Naive Bayes algorithm 2.4242 2.8425 Trilateration 7.3085 3.1542 Transfer learning algorithm 2.0594 1.7969

[0046] Table 2 Prediction error and variance of each indoor positioning algorithm in the laboratory

[0047] algorithm Prediction error (m) <![CDATA[Variance (m 2 )]]> Kernel adaptive filter of the present invention 1.1213 0.3448 KRLS 1.3382 0.7900 KLMP 1.7450 1.1290 KLMS 1.7124 1.6139 KRLP 1.3649 0.6539 KRMC 1.3573 0.7568 K-nearest neighbor algorithm 1.3856 0.5921 Naive Bayes algorithm 2.8274 3.5859 Trilateration 4.7911 1.4571 Transfer learning algorithm 2.0594 1.7969

Claims

1. An indoor positioning method using adaptive data processing, the method comprising: Step 1: Offline learning phase: The actual position of the movable sensor in the target area is known. i The RSSI value is measured at the reference point to form the input RSSI data ui; Step 2: Use kernel adaptive filter based on input RSSI data u i Get the estimated position y i , by estimating the position y i With the actual position d i Comparison to obtain the position error e i , and then modify the internal weights of the kernel adaptive filter to achieve the training data sequence {u i ,d i }'s learning; Step 3: Online application stage: Actual location d i Unknown input RSSI data u i is passed into the fully trained kernel adaptive filter, which gives the estimated position y i , the estimated position y i As the input RSSI data u i The position estimation result is directly output.

2. The indoor positioning method of adaptive data processing according to claim 1, characterized in that: The specific method of the kernel adaptive filter is: Step 1: Define a sequence sample set {u(i), d(i)}, where u(i) represents the input RSSI data at time i to be filtered, and d(i) represents the corresponding actual position; define the kernel function k σ (·), kernel function k σ (·) can map the input RSSI data u(i) into a high-dimensional feature space In this paper, we use data processing in high-dimensional space to learn the nonlinear relationship of the input RSSI data u(i); define The high-dimensional input RSSI data u(i) is mapped to the Hilbert space H based on the kernel function; the high-dimensional input RSSI data The relationship with the low-dimensional input RSSI data u(i) is: Step 2: Define the loss function L of the kernel adaptive filter as: Among them, ω represents the weight of the kernel adaptive filter, G θ (·) is the Gaussian kernel function, is the position error at time i; λ is the regularization parameter, β is the forgetting factor, λ‖ω j ‖ 2 is the regularization term, q represents the value of the series, ω i represents the updated weight of the kernel adaptive filter at time i, Represents the high-dimensional input RSSI data u(i) mapped to the Hilbert space H based on the kernel function; Step 3: Define the updated weights ω of the kernel adaptive filter i : Among them, Φ i The expansion of the input RSSI data u(i) High-dimensional input RSSI data obtained by mapping in Hilbert space H; ω i To update the weight; Step 4: According to the kernel function shown in formula (1), the weights of the modified kernel adaptive filter are continuously updated according to the input RSSI data u(i) and the actual position d(i) to achieve offline learning.

3. The indoor positioning method of adaptive data processing according to claim 1, characterized in that: In the online application phase of step 3, based on the input RSSI data u(i) whose actual position d(i) is unknown, the estimated position y(i) is obtained by the following relationship: Subsequently, y(i) is directly output as the position estimate of the input RSSI data of the unknown position by the positioning system to realize the positioning function.

4. The indoor positioning method of adaptive data processing according to claim 2, characterized in that: The kernel adaptive filter is a q-order kernel adaptive filter.

Images