Indoor positioning method and system based on dynamic mode switching and data fusion

Abstract

The invention provides an indoor positioning method and system based on dynamic mode switching and data fusion, and relates to the technical field of visible light indoor positioning, and the method comprises the following steps: deploying a VLC transmitting device and an auxiliary positioning point, constructing a VLC fingerprint database, carrying out dynamic signal detection and mode conversion, carrying out VLC fingerprint positioning, carrying out IMU data processing, and carrying out main positioning mode data fusion. VLC signal intensity distance estimation, ultrasonic ranging, IMU relative displacement estimation and standby positioning mode data fusion are carried out. According to the invention, by deploying the visible light communication transmitting device and the auxiliary positioning point, the strength and the stability of the VLC signal are monitored in real time, and dynamic positioning mode switching is realized through fuzzy logic reasoning. In the main positioning mode, VLC fingerprint matching is combined with IMU data, and accurate positioning is carried out through a Kalman filtering algorithm; and when the VLC signal is weak or unstable, a standby positioning mode is switched, and accurate positioning is carried out by fusing data through a particle filter algorithm according to VLC signal intensity distance estimation, ultrasonic ranging and I MU data.

Description

Technical Field

[0001] The present invention relates to the field of indoor positioning technology, and in particular to an indoor positioning method and system based on dynamic mode switching and data fusion. Background Art

[0002] With the rapid development of informatization and intelligence, the demand for indoor positioning technology is increasing. The traditional global positioning system (GPS) is limited by signal attenuation and multipath effects in indoor environments, and there is an urgent need to develop efficient and accurate indoor positioning solutions. Visible light communication (VLC) is an emerging indoor positioning technology that uses LED lights as signal sources and modulates optical signals for data transmission. VLC signals have high frequency and large bandwidth, can achieve fast positioning, and are not subject to electromagnetic interference. They are suitable for places with strict requirements on electromagnetic compatibility.

[0003] Named "An Indoor Positioning Detection System and Method", the method first deploys LED lights in a certain layout, and sets up a three-dimensional coordinate system to collect the signal strength data received by the receiver; based on the relative position and angle of the LED and the receiver, a mathematical model of the visible light channel is established to calculate the signal strength data under the direct line of sight path and one reflection; the signal strength and coordinate data are collected as training data sets, and a machine learning model is built for training; based on experiments with different numbers and layouts of LED combinations, the number and layout of LEDs that minimize the positioning error are selected to improve the positioning accuracy; the trained machine learning model is used to predict the position of the photoelectric receiver, and the positioning accuracy is calculated by the mean square error. This method has a certain improvement in positioning accuracy, but does not consider the response strategy when the VLC signal is interfered or the light source is unstable, and does not involve the functions of multi-label recognition and parallel positioning.

[0004] Therefore, the development of indoor positioning methods and systems based on dynamic mode switching and data fusion can flexibly utilize VLC and other sensor data to form a more efficient indoor positioning solution, which has important theoretical and practical application value. Summary of the invention

[0005] The technical solution of the present invention to solve the above technical problems is to provide an indoor positioning method based on dynamic mode switching and data fusion, comprising the following steps:

[0006] Step 1, deploy VLC transmitting equipment and auxiliary positioning points: establish a Cartesian coordinate system in the environment to be positioned; deploy VLC transmitting equipment on the ceiling in a regular network layout, and calibrate the coordinates of each VLC light source; set auxiliary positioning points and calibrate their coordinates;

[0007] Step 2, constructing a VLC fingerprint database: collecting signal feature data of each LED light source through multiple signal collection points with known coordinates, including RSSI, phase information and arrival time, and performing low-pass filtering on the collected signal feature values ​​to form representative signal feature values ​​and store them in the database;

[0008] Step 3, dynamic signal detection and mode conversion: Real-time detection of VLC signal reception strength and stability, using fuzzy logic rules to dynamically select the main positioning mode or the backup positioning mode according to signal strength and stability, or enable both modes at the same time and perform weighted fusion of the positioning results;

[0009] Step 4, VLC fingerprint positioning: In the main positioning mode, the VLC receiver of the tag to be positioned collects signal features in real time, performs preprocessing and feature extraction, matches the data in the fingerprint database, and uses the nearest neighbor algorithm to obtain the estimated position;

[0010] Step 5, IMU data processing: In the main positioning mode, the IMU data of the tag is collected in real time, and the IMU data is pre-processed to analyze the dynamic motion information of the tag;

[0011] Step 6, main positioning mode data fusion: using the Kalman filter model, VLC fingerprint positioning results and IMU data are combined to obtain accurate three-dimensional coordinates;

[0012] Step 7, VLC signal strength distance estimation: In the standby positioning mode, the distance between the tag and the VLC transmitter is calculated based on the RSSI of the VLC signal and the path loss model, and the distance is corrected using the phase information;

[0013] Step 8: Ultrasonic ranging: In the standby positioning mode, the round trip time of the ultrasonic signal is used to calculate the distance between the tag and the auxiliary positioning point;

[0014] Step 9, IMU relative displacement estimation: In the backup positioning mode, the relative displacement is obtained by double integration of IMU data;

[0015] Step 10, backup positioning mode data fusion: multi-modal fusion of VLC, IMU and ultrasonic data is realized through particle filter algorithm to obtain the final position estimate of the tag.

[0016] Furthermore, the fuzzy logic rules in step 3 include:

[0017] Rule 1: If the VLC signal is strong and stable, enable the main positioning mode;

[0018] Rule 2: If the VLC signal is weak and unstable, enable the backup positioning mode;

[0019] Rule 3: If the VLC signal strength is medium and the signal stability is medium stable, both the main positioning mode and the backup positioning mode are enabled and the positioning results of the two are weightedly fused.

[0020] Furthermore, the Kalman filter model in step 6 includes:

[0021] Define the state variables of the tag, including position coordinates and speed, to form a state vector;

[0022] Use the three-axis acceleration in the IMU data as control input to update the velocity and position;

[0023] Establish the state transfer equation based on IMU data, and use IMU data to predict the state of the tag position through Kalman filtering;

[0024] Combined with the VLC fingerprint positioning results, the position is corrected and the state vector and error covariance matrix of the tag are updated.

[0025] Furthermore, the particle filter algorithm in step 10 includes:

[0026] Initialize the particle filter and randomly generate particles within the known initial position range. Each particle represents the possible position of the label.

[0027] State prediction based on IMU data, applying the motion information generated by IMU data to each particle to simulate the possible movement path of the tag;

[0028] Multimodal observation update of VLC and ultrasonic data, using RSSI, phase information of VLC signal and ultrasonic ranging value to update particle weights;

[0029] Particle resampling: resample particles based on weights, retain high-weight particles, generate new particles near these particles, and abandon low-weight particles;

[0030] Position output and dynamic update: The final position output obtained by weighted average calculation is continuously integrated with new data to achieve dynamic tracking of tag position.

[0031] In order to solve the above technical problems, the present invention further proposes an indoor positioning system based on dynamic mode switching and data fusion, which is used to execute the indoor positioning method based on dynamic mode switching and data fusion as described above, comprising:

[0032] Data acquisition module, including VLC signal acquisition submodule, IMU data acquisition submodule and ultrasonic ranging submodule;

[0033] Data preprocessing module, including signal denoising submodule and feature extraction submodule;

[0034] The signal quality monitoring and mode switching module includes a VLC signal detection submodule and a mode switching control submodule, which is used to monitor the VLC signal strength and stability in real time and select the positioning mode according to fuzzy logic rules;

[0035] Data fusion module, including Kalman filter algorithm and particle filter algorithm, used to fuse VLC, IMU and ultrasonic data;

[0036] Position estimation and output module, used to output the tag position information obtained after data fusion in real time;

[0037] The system control module is used to manage the cyclic update operation of the system and control the execution order of modules such as data acquisition, preprocessing, and fusion.

[0038] Furthermore, the design of the fuzzy logic rules takes into account the membership of signal strength and stability, as well as the weight distribution of different positioning modes, so as to improve the adaptability and robustness of the system.

[0039] Furthermore, the particle filter algorithm updates particle weights through multimodal observations and uses resampling technology to reduce positioning errors, thereby achieving high-precision positioning under long-term operation.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. The present invention provides a fuzzy logic switching and fusion positioning strategy for visible light indoor positioning. By implementing regularized reasoning of continuous signals through fuzzy logic, the positioning mode can be dynamically switched according to the strength and stability of the VLC signal in the actual environment, avoiding the problem of frequent switching of the positioning mode at the critical point and enhancing stability. When the signal is medium, the positioning accuracy is further improved by weighted fusion positioning mode results.

[0042] 2. In the main positioning mode, the present invention uses the Kalman filter algorithm to combine VLC and IMU data to predict the motion trajectory, and in the auxiliary positioning mode, the particle filter algorithm is used to fuse VLC data, IMU data and ultrasonic data, so as to correct the tag position in real time, and the positioning error can be controlled within the range of 5-10 cm on average.

[0043] 3. The present invention provides a system that supports multi-tag identification and parallel positioning. Through the independent ID data of each tag, the present invention can realize the identity recognition and location determination of multiple tags. The system can realize the parallel positioning of multiple targets based on the ID of each independent tag, meeting the actual needs of multi-tag positioning scenarios such as logistics and warehousing. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0045] Figure 1 It is a flowchart of the steps of the indoor positioning method based on dynamic mode switching and data fusion of the present invention;

[0046] Figure 2 It is a schematic diagram of the layout model of the present invention;

[0047] Figure 3 It is a structural diagram of the indoor positioning system based on dynamic mode switching and data fusion of the present invention.

[0048] Description of reference numerals:

[0049] 1. Data acquisition module, 2. Data preprocessing module, 3. Signal quality monitoring and mode switching module, 4. Data fusion module, 5. Position estimation and output module, 6. System control module. DETAILED DESCRIPTION

[0050] The present invention proposes an indoor positioning method and system based on dynamic mode switching and data fusion, aiming to design an indoor positioning method based on dynamic mode switching and data fusion to achieve higher positioning accuracy, stability and dynamic environment adaptability.

[0051] The indoor positioning method and system based on dynamic mode switching and data fusion proposed by the present invention will be described in the following specific embodiments:

[0052] Embodiment 1:

[0053] In the technical solution of this embodiment, Figure 1 As shown, an indoor positioning method based on dynamic mode switching and data fusion includes the following steps:

[0054] Step 1, deploy VLC transmitting equipment and auxiliary positioning points: establish a Cartesian coordinate system in the environment to be positioned; deploy VLC transmitting equipment on the ceiling in a regular network layout, and calibrate the coordinates of each VLC light source; set auxiliary positioning points and calibrate their coordinates;

[0055] Specifically, the layout of LED light sources and auxiliary positioning points refers to Figure 2As shown. In the indoor environment to be positioned, a Cartesian coordinate system is established with point O as the origin, and high-power LED light sources are deployed in a regular grid on the ceiling. According to the divergence angle and coverage of the high-power LED, the distance between the transmitting devices can be set to 5 meters, and the three-dimensional coordinates of each LED light source are calibrated; in areas where the VLC signal may be weak or susceptible to interference, fixed auxiliary positioning points are arranged, and the auxiliary positioning points integrate independent VLC transmitters and ultrasonic sensors; the tag to be positioned integrates a VLC receiver for receiving VLC signals, and an integrated IMU sensor for recording the acceleration and other motion information of the tag, providing auxiliary support for the continuous positioning of the tag.

[0056] Step 2, constructing a VLC fingerprint database: collecting signal feature data of each LED light source through multiple signal collection points with known coordinates, including RSSI, phase information and arrival time, and performing low-pass filtering on the collected signal feature values ​​to form representative signal feature values ​​and store them in the database;

[0057] Specifically, first, each LED light source carries a unique signal feature when emitting a light signal, and the VLC signal receiving end on the tag can be used to identify the specific location of the light source. Secondly, the VLC fingerprint data is collected, and multiple known coordinate points are selected as signal collection points. The collection points should be distributed as evenly as possible in the indoor space to ensure the comprehensiveness of the fingerprint database data. At each collection point, a VLC signal receiver with multiple tags is used multiple times to collect the signal feature data of each LED light source, including RSSI, phase information, and arrival time (ToA). Furthermore, the signal feature value collected each time will be filtered out of noise through a low-pass filter, and finally a representative signal feature value of the collection point will be formed, and the processed fingerprint data will be stored in the database.

[0058] Step 3, dynamic signal detection and mode conversion: real-time detection of VLC signal reception strength and stability, using fuzzy logic rules to dynamically select the main positioning mode, the backup positioning mode, or both modes according to signal strength and stability, and weighted fusion of positioning results;

[0059] Specifically, the dynamic signal detection and mode conversion process: First, real-time dynamic signal monitoring and fuzzy processing are performed. The VLC signal reception strength S and its stability are detected in real time, and the short-time standard deviation σ is used to measure the signal fluctuation:

[0060]

[0061] in, is the mean value of RSSI, and N is the number of sampling points.

[0062] Use triangular membership functions to partition the fuzzy sets of signal strength:

[0063]

[0064] Among them, S min is the minimum value of signal strength; S mid is the median value of signal strength; S max is the maximum value of signal strength.

[0065] Partition stability of fuzzy sets:

[0066]

[0067] Among them, σ min is the minimum value of signal stability; σ mid is the middle value of signal stability; σ max is the maximum value of signal stability.

[0068] Secondly, according to the signal strength and stability, set the fuzzy logic rules:

[0069] Rule 1: If the signal is strong and stable, enable the main positioning mode and perform steps 4 to 6.

[0070] Rule 2: If the signal is weak and unstable, enable the backup positioning mode and perform steps 7 to 10.

[0071] Rule 3: If the signal strength is medium and the signal stability is medium stable, both the VLC mode and the backup mode are enabled and the positioning results of the two are weighted fused.

[0072] Use the membership function to calculate the matching degree of each rule:

[0073] μ 规则1 =min(μ 强 (S),μ 稳定 (σ));

[0074] μ 规则2 =min(μ 弱 (S),μ 不稳定 (σ));

[0075] μ 规则3 =min(μ 中等 (S),μ 中等稳定 (σ));

[0076] The membership of each positioning mode is calculated using the weighted average method:

[0077]

[0078] Among them, μ 模式 is the membership degree of the final positioning mode; μ 规则i is the matching degree of the i-th rule; wi is the weight of the ith rule; N is the number of all relevant rules.

[0079] Third, according to the result and membership of fuzzy reasoning, select the current positioning mode. If the membership of the main positioning mode is higher than that of the backup positioning mode, the main positioning mode is enabled; if the membership of the backup positioning mode is higher than that of the main positioning mode, the backup positioning mode is enabled; when the signal strength and stability are moderate, use both positioning modes for positioning, and perform weighted fusion of the positioning results according to the membership of the two modes:

[0080] P 融合 =w 主 ·P 主 +w 备用 ·P 备用 ;

[0081] Among them, P 融合 is the final positioning result in the fusion mode; P 主 and P 备用 are the positioning results of the main positioning mode and the backup positioning mode respectively; w 主 and w 备用 is the weight obtained from the membership degree:

[0082]

[0083] Step 4, VLC fingerprint positioning: In the main positioning mode, the VLC receiver of the tag to be positioned collects signal features in real time, performs preprocessing and feature extraction, matches the data in the fingerprint database, and uses the nearest neighbor algorithm to obtain the estimated position;

[0084] Specifically, in the main positioning mode, first, the VLC receiver of the tag to be positioned collects signal characteristics (RSSI, phase, ToA) in real time.

[0085] Secondly, pre-process the collected real-time data. Perform low-pass filtering on RSSI, phase and arrival time data to remove random noise. Taking RSSI as an example, the low-pass filtering formula is as follows:

[0086]

[0087] in, is the RSSI value at the current time t after filtering; α is the filter coefficient, which is used to smooth RSSI fluctuations; S RSSI (t) is the original RSSI value at time t, i.e., the instantaneous RSSI data measured directly from the receiver; is the RSSI value after filtering at the previous time point t-1.

[0088] Again, extract key features from the denoised data: RSSI average, phase principal component information, arrival time average, etc. RSSI mean Used to describe the distribution of signal strength and can effectively characterize distance; Phase principal component information PC Can help distinguish multipath signals; mean of arrival time Used for distance estimation.

[0089] Again, the extracted real-time signal features are matched with the data in the fingerprint database, and the nearest neighbor (k-NN) algorithm is used to find the coordinates closest to the current feature. Assume that the real-time feature vector is By calculating the current feature vector and the fingerprint database feature vector f i The Euclidean distance is:

[0090]

[0091] Calculate the distance between the real-time data and each sample in the database to form a distance set {d1, d2, ..., d m}, find the sample with the smallest distance based on the distance set.

[0092] Matching result x VLC =(x VLC ,y VLC ,z VLC ) is the estimated position of the tag to be located obtained by VLC fingerprint matching.

[0093] Step 5, IMU data processing: In the main positioning mode, the IMU data of the tag is collected in real time, and the IMU data is pre-processed to analyze the dynamic motion information of the tag;

[0094] Specifically, in the main positioning mode, the IMU sensor integrated on each tag to be positioned collects the tag’s three-axis acceleration (a x ,a y ,a z ) and other directional information. The collected IMU data is used to record the dynamic motion state of the tag, including motion direction, acceleration change, etc., to provide support for subsequent positioning. The IMU data is filtered and smoothed, and low-pass filtering is used to remove high-frequency noise.

[0095] Step 6, main positioning mode data fusion: using the Kalman filter model, VLC fingerprint positioning results and IMU data are combined to obtain accurate three-dimensional coordinates;

[0096] Specifically, based on the estimated position obtained in step 4 and the IMU data collected in step 5, a Kalman filter model is constructed. The state variables of the tag are defined, including the position coordinates and velocity, to form a state vector x k :

[0097]

[0098] Among them, x k ,y k and z k Indicates the three-dimensional position coordinates of the label, v xk 、v yk and v zk Represents the velocity components of the tag in the x, y, and z directions.

[0099] Using the three-axis acceleration (a x ,a y ,a z ) as control input to update velocity and position.

[0100] Assuming that the movement of the tag conforms to the uniform speed or uniform acceleration model, the state transfer equation is established based on the IMU data:

[0101] x k =A·x k-1 +B·U k-1 +w k-1 ;

[0102] Among them, x k-1 is the state vector at the previous moment. The control input vector is w k-1 is the process noise, which indicates the random error caused by the dynamic changes of the system. The state transfer matrix A represents the dynamic model of the system. Under uniform motion, A can be expressed as:

[0103]

[0104] Where Δt is the time step. The control input matrix B is combined with the acceleration of the IMU (a x ,a y ,a z ) Update the velocity component to control the state update:

[0105]

[0106] Establish the observation vector z k To update the Kalman filter state estimate:

[0107] z k =H·x k +v k ;

[0108] Among them, v k is the observation noise, which represents the random error of VLC fingerprint positioning measurement, and H is the observation matrix, which is used to extract the position information in the state vector:

[0109]

[0110] Secondly, IMU data drives state prediction. The IMU data is used to predict the state of the tag position through Kalman filtering. The three-dimensional position and velocity of the tag at the next moment are predicted using the state transition equation and control input:

[0111]

[0112] in, Represents the prediction of the label state at time k.

[0113] Update the prediction error covariance matrix based on the process noise Q:

[0114] P k|k-1 =A·P k-1|k-1 ·A T +Q;

[0115] Among them, P k|k-1 is the prediction error covariance matrix, which indicates the reliability of the current prediction.

[0116] Third, the position is corrected by combining the VLC fingerprint positioning result. Calculate the Kalman gain K k , used to balance IMU predictions and VLC measurement data:

[0117] K k =P k|k-1 ·H T ·(H·P k|k-1 ·H T +R) -1 ;

[0118] Among them, R is the observation noise covariance matrix, which reflects the uncertainty of VLC positioning.

[0119] Use the Kalman gain to correct the position of the IMU predicted state and update the state vector of the label:

[0120]

[0121] in, Represents the difference between VLC observation and IMU prediction.

[0122] Update the error covariance matrix to ensure that the prediction error at the next moment is optimized and the error is gradually reduced over time:

[0123] P k|k =(Ι-K k ·H)·P k|k-1 ;

[0124] Where Ι represents the identity matrix.

[0125] After Kalman filter fusion, the final state vector [x k y k z k ] T That is, the precise three-dimensional coordinates after fusion in the main positioning mode.

[0126] Step 7, VLC signal strength distance estimation: In the standby positioning mode, the distance between the tag and the VLC transmitter is calculated based on the RSSI of the VLC signal and the path loss model, and the distance is corrected using the phase information;

[0127] Specifically, in the standby positioning mode, the VLC transmitter of the auxiliary positioning point broadcasts an optical signal that carries specific coded information so that the system can identify the source of the signal, and the tag receiver measures the received signal strength S RSSI . Calculate the distance d between VLC and the tag based on the path loss model VLC :

[0128]

[0129] Wherein, d0 is the reference distance, S0 is the reference signal strength, and n is the path loss factor.

[0130] Phase information is used to further correct the distance. Phase information reflects the position angle of the tag relative to the transmitter and can be used to improve the position prediction. Assume that the tag can receive the phase information φ of the two transmitters. i and φ j , then the phase difference Δφ between the two can be expressed as:

[0131] Δφ=φ i -φ j ;

[0132] Among them, φ i and φ j are the signal phases received from VLC transmitters i and j, respectively.

[0133] Step 8: Ultrasonic ranging: In the standby positioning mode, the round trip time of the ultrasonic signal is used to calculate the distance between the tag and the auxiliary positioning point;

[0134] Specifically, in the standby positioning mode, the ultrasonic signal is emitted and reflected by the tag, and the auxiliary positioning point calculates the distance d according to the round-trip time Δt of the ultrasonic wave. US :

[0135]

[0136] Among them, c ultrasonic is the propagation speed of ultrasound in the air. The ranging data provides the precise distance information between the tag and the auxiliary positioning point.

[0137] Step 9, IMU relative displacement estimation: In the backup positioning mode, the relative displacement is obtained by double integration of IMU data;

[0138] Specifically, in the standby positioning mode, the acceleration data obtained by the IMU is preprocessed and filtered, and the relative displacement is obtained by double integration:

[0139] First, calculate the speed:

[0140] v t =v t-1 +a t-1 Δt;

[0141] Among them, v t is the velocity vector at the current time t; a t-1 is the acceleration at the previous moment; Δt is the time interval.

[0142] On this basis, the position change is calculated:

[0143]

[0144] Among them, x t is the position at the current time t; x t-1 The position at the previous moment.

[0145] Step 10, backup positioning mode data fusion: multi-modal fusion of VLC, IMU and ultrasonic data is realized through particle filter algorithm to obtain the final position estimate of the tag.

[0146] Specifically, the multimodal fusion of VLC, IMU and ultrasonic data is realized through the particle filter algorithm.

[0147] First, initialize the particle filter.

[0148] N particles are randomly generated within the known initial position range, and each particle represents the possible position of the label. Each particle contains three-dimensional position and velocity information, and the state vector of particle i is defined as:

[0149]

[0150] in, is the two-dimensional coordinate of particle i at time t. When the initial state is uncertain, uniform distribution can be used.

[0151] Assign the same initial weight to each particle i Indicates that each position hypothesis is equally likely at the beginning of positioning, that is:

[0152]

[0153] Second, state prediction based on IMU data

[0154] According to the tag displacement obtained in step 9, the motion information generated by the IMU data is applied to each particle to simulate the possible movement path of the tag. The state update formula is as follows:

[0155]

[0156] in, is the position of particle i at time t+1; is the current position of particle i at time t, is the velocity of particle i, usually Obtained by integrating the previous moment velocity and the current IMU acceleration, is the acceleration of particle i measured by IMU at the current moment, is the process noise, which obeys the zero-mean Gaussian distribution to simulate uncertainty, and Δt is the time step.

[0157] Third, multimodal observations of VLC and ultrasound data are updated.

[0158] Use the RSSI and phase information of the VLC signal obtained in step 7 and the ultrasonic ranging value obtained in step 8 to update the particle weight, estimate the current position through the measurement data, and then convert the matching degree between the estimated position and the measurement value of each particle into weight update:

[0159]

[0160] Among them, d VLC is the measured distance estimated based on the RSSI value, indicating the distance from the tag to the VLC transmitting device; is the predicted RSSI distance of particle i; Δφ is the phase difference of the VLC signal, indicating the relative position between the tag and multiple VLC transmitters; Δφ (i) is the predicted phase difference of particle i; d US The distance between the tag and the auxiliary positioning point obtained by ultrasonic measurement; is the predicted ultrasonic distance of particle i; σ VLC , σ φ and σ US are the standard deviations of RSSI, phase, and ultrasonic measurements, respectively.

[0161] Updated weights Indicates the degree of match between the particle and the measured data. The larger the weight, the higher the degree of match between the particle's position and the observed data.

[0162] Fourth, particle resampling.

[0163] Resample the particles based on the weights, retain the particles with high weights, generate new particles near these particles, and abandon the particles with lower weights. This step gradually concentrates the particle swarm near the true position and reduces the positioning error. Calculate the weighted average position of all particles as the final position estimate of the label:

[0164]

[0165] Among them, the weighted average position estimate is an accurate estimate of the tag's current position.

[0166] Fifth, position output and dynamic update.

[0167] The final position calculated by weighted average The output is used as the positioning result of the tag at the current moment for real-time display by the upper-level system or application. Over time, the above prediction, observation update and resampling steps are repeated to continuously integrate new IMU, VLC and ultrasonic data, adjust the position distribution of particles, and realize dynamic tracking of tag positions. When the signal conditions are good, the system gives priority to using VLC data to give particles a higher weight; when the VLC signal is weak, the weight of IMU and ultrasonic data is automatically increased to ensure the stability of the system under signal fluctuations.

[0168] Furthermore, the fuzzy logic rules in step 3 include:

[0169] Rule 1: If the VLC signal is strong and stable, enable the main positioning mode;

[0170] Rule 2: If the VLC signal is weak and unstable, enable the backup positioning mode;

[0171] Rule 3: If the VLC signal strength is medium and the signal stability is medium stable, both the VLC mode and the backup mode are enabled and the positioning results of the two are weightedly fused.

[0172] Furthermore, the Kalman filter model in step 6 includes:

[0173] Define the state variables of the tag, including position coordinates and speed, to form a state vector;

[0174] Use the three-axis acceleration in the IMU data as control input to update the velocity and position;

[0175] Establish the state transfer equation based on IMU data, and use IMU data to predict the state of the tag position through Kalman filtering;

[0176] Combined with the VLC fingerprint positioning results, the position is corrected and the state vector and error covariance matrix of the tag are updated.

[0177] Furthermore, the particle filter algorithm in step 10 includes:

[0178] Initialize the particle filter and randomly generate particles within the known initial position range. Each particle represents the possible position of the label.

[0179] State prediction based on IMU data, applying the motion information generated by IMU data to each particle to simulate the possible movement path of the tag;

[0180] Multimodal observation update of VLC and ultrasonic data, using RSSI, phase information of VLC signal and ultrasonic ranging value to update particle weights;

[0181] Particle resampling: resample particles based on weights, retain high-weight particles, generate new particles near these particles, and abandon particles with lower weights;

[0182] Position output and dynamic update: The final position output is calculated by weighted average, and new data is continuously integrated over time to achieve dynamic tracking of the tag position.

[0183] Embodiment 2:

[0184] An indoor positioning system based on dynamic mode switching and data fusion is used to execute the indoor positioning method based on dynamic mode switching and data fusion as described in Example 1. The system can realize dynamic positioning mode switching through fuzzy logic reasoning and signal feature analysis. Multimodal data fusion based on visible light communication (VLC), inertial measurement unit (IMU) and ultrasonic sensor realizes high-precision indoor positioning. The system responds to the multi-tag positioning requirements in complex indoor environments through dynamic signal detection, positioning mode conversion, data fusion and real-time update of particle filter algorithm. The system hardware structure includes a VLC signal transmitter, an auxiliary positioning point and a tag to be positioned. For the VLC signal transmitter, the system arranges multiple LED light sources in the indoor environment to be positioned as VLC signal transmitters. The LED light sources are distributed in key positions in the positioning area to form an effective signal coverage network. Each light source carries a unique signal feature, including information such as signal strength (RSSI) and phase (Phase), so that the tag receiver can distinguish different light sources and determine their relative positions. For auxiliary positioning points, auxiliary positioning points are arranged in areas where the VLC signal may be weak. Each auxiliary positioning point is equipped with an independent VLC transmitter and ultrasonic sensor. For the tag to be located, a VLC signal receiver, an IMU (inertial measurement unit) and a module for reflecting ultrasonic signals are integrated.

[0185] Indoor positioning systems include:

[0186] Data acquisition module, including VLC signal acquisition submodule, IMU data acquisition submodule and ultrasonic ranging submodule;

[0187] Specifically, the VLC signal acquisition submodule is responsible for receiving and collecting VLC signal feature data such as RSSI, phase and other data; the IMU data acquisition submodule collects acceleration motion data from the IMU on the tag; the ultrasonic ranging submodule is responsible for obtaining the ultrasonic ranging data between the tag and the positioning point from the auxiliary positioning point.

[0188] Data preprocessing module, including signal denoising submodule and feature extraction submodule;

[0189] Specifically, the signal denoising submodule performs low-pass filtering and denoising on the VLC, IMU and ultrasonic data respectively to reduce the impact of environmental interference; the feature extraction submodule extracts key features such as RSSI average, phase information, distance measurement, etc. from the denoised data to provide clear data features for subsequent fusion calculations.

[0190] The signal quality monitoring and mode switching module includes a VLC signal detection submodule and a mode switching control submodule, which is used to monitor the VLC signal strength and stability in real time and select the positioning mode according to fuzzy logic rules;

[0191] Specifically, the VLC signal detection submodule is responsible for monitoring the strength and stability of the VLC signal, and triggering auxiliary positioning when the signal is poor or unstable; the mode switching control submodule uses the membership function to calculate the matching degree of each rule and the membership of each positioning mode according to the fuzzy logic rules, and selects the current positioning mode according to the results of fuzzy reasoning and the membership to ensure the adaptability of the system.

[0192] Data fusion module, including Kalman filter algorithm and particle filter algorithm, used to fuse VLC, IMU and ultrasonic data;

[0193] Specifically, Kalman filtering is used to fuse the VLC data and IMU data in the main positioning mode; based on the particle filtering algorithm, the VLC signal, ultrasonic ranging data and IMU data are fused and calculated.

[0194] The position estimation and output module is used to output the tag position information obtained after data fusion in real time and provide it to external devices or upper-level applications.

[0195] The system control module is used to manage the cyclic update operation of the system and control the execution order of modules such as data acquisition, preprocessing, and fusion.

[0196] When the signal is sufficiently stable, the system uses the VLC fingerprint database to achieve accurate positioning; when the VLC signal is poor, it can use auxiliary positioning points and IMU predictions to maintain the continuity of positioning. The particle filter algorithm in the backup positioning mode further reduces the impact of IMU drift, allowing the system to adapt to dynamic and complex indoor environments and meet the high-precision positioning requirements of multiple tags.

[0197] The indoor positioning system based on dynamic mode switching and data fusion combines the data fusion technology of VLC, IMU and ultrasonic ranging, and has significant advantages, which are specifically reflected in positioning accuracy, stability, environmental adaptability and real-time performance. Compared with traditional single positioning technology, the present invention shows higher performance and reliability in complex indoor environments.

[0198] Dynamic signal detection and mode switching: The system performs dynamic signal detection based on real-time signal strength and stability, and uses fuzzy logic control for dynamic mode switching. Fuzzy logic dynamically adjusts the positioning mode through the membership of signal strength and stability, avoiding the "frequent jitter" problem in threshold switching. In the case of traditional threshold switching, the switching frequency is 10 times per minute when the signal boundary jitters, causing the positioning accuracy to drop to about 85%. Under the same circumstances, using fuzzy logic switching, the frequency is reduced to 2 times per minute, and the positioning accuracy remains above 93%. Fuzzy logic can reduce the impact of signal jitter on mode switching and improve robustness.

[0199] Particle filter suppresses IMU drift: By processing IMU data with a particle filter algorithm, this method effectively suppresses the drift error of the IMU during long-term operation. The IMU can provide accurate motion data in a short period of time, but it is easy to accumulate drift over time. By fusing VLC and ultrasonic data for real-time correction, this system can minimize the drift error. Under long-term operation conditions (>1 hour), the IMU drift error increase of this method is less than 5 cm, while the uncorrected IMU drift error can reach more than 30 cm. Multimodal fusion greatly reduces drift accumulation and ensures long-term positioning accuracy.

[0200] VLC fingerprint database supports multi-tag identification: The VLC fingerprint database used by the system can distinguish the signal characteristics of different tags, and realize the identification and parallel positioning of multiple tags. Compared with the traditional single-tag positioning method, this method has significant advantages in logistics, warehousing and other scenarios that require tracking multiple mobile tags. In a warehousing environment, 10 tags are positioned simultaneously, and the average positioning delay of this method is less than 500 milliseconds, and the positioning accuracy is maintained within 5-10 centimeters. However, in traditional single-mode positioning methods, the positioning error and delay are significantly increased in multi-tag scenarios.

[0201] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. An indoor positioning method based on dynamic mode switching and data fusion, characterized in that: The following steps are involved: Step 1, deploy VLC transmitting equipment and auxiliary positioning points: establish a Cartesian coordinate system in the environment to be positioned; deploy VLC transmitting equipment on the ceiling in a regular network layout, and calibrate the coordinates of each VLC light source; set auxiliary positioning points and calibrate their coordinates; Step 2, constructing a VLC fingerprint database: collecting signal feature data of each LED light source through multiple signal collection points with known coordinates, including RSSI, phase information and arrival time, and performing low-pass filtering on the collected signal feature values ​​to form representative signal feature values ​​and store them in the database; Step 3, dynamic signal detection and mode conversion: Real-time detection of VLC signal reception strength and stability, using fuzzy logic rules to dynamically select the main positioning mode or the backup positioning mode according to signal strength and stability, or enable both modes at the same time and perform weighted fusion of the positioning results; Step 4, VLC fingerprint positioning: In the main positioning mode, the VLC receiver of the tag to be positioned collects signal features in real time, performs preprocessing and feature extraction, matches the data in the fingerprint database, and uses the nearest neighbor algorithm to obtain the estimated position; Step 5, IMU data processing: In the main positioning mode, the IMU data of the tag is collected in real time, and the IMU data is pre-processed to analyze the dynamic motion information of the tag; Step 6, main positioning mode data fusion: using the Kalman filter model, VLC fingerprint positioning results and IMU data are combined to obtain accurate three-dimensional coordinates; Step 7, VLC signal strength distance estimation: In the standby positioning mode, the distance between the tag and the VLC transmitter is calculated based on the RSSI of the VLC signal and the path loss model, and the distance is corrected using the phase information; Step 8: Ultrasonic ranging: In the standby positioning mode, the round trip time of the ultrasonic signal is used to calculate the distance between the tag and the auxiliary positioning point; Step 9, IMU relative displacement estimation: In the backup positioning mode, the relative displacement is obtained by double integration of IMU data; Step 10, backup positioning mode data fusion: multi-modal fusion of VLC, IMU and ultrasonic data is realized through particle filter algorithm to obtain the final position estimate of the tag.

2. The indoor positioning method based on dynamic mode switching and data fusion according to claim 1, characterized in that: The fuzzy logic rules in step 3 include: Rule 1: If the VLC signal is strong and stable, enable the main positioning mode; Rule 2: If the VLC signal is weak and unstable, enable the backup positioning mode; Rule 3: If the VLC signal strength is medium and the signal stability is medium stable, both the main positioning mode and the backup positioning mode are enabled and the positioning results of the two are weightedly fused.

3. The indoor positioning method based on dynamic mode switching and data fusion according to claim 1, characterized in that: The Kalman filter model in step 6 includes: Define the state variables of the tag, including position coordinates and speed, to form a state vector; Use the three-axis acceleration in the IMU data as control input to update the velocity and position; Establish the state transfer equation based on IMU data, and use IMU data to predict the state of the tag position through Kalman filtering; Combined with the VLC fingerprint positioning results, the position is corrected and the state vector and error covariance matrix of the tag are updated.

4. The indoor positioning method based on dynamic mode switching and data fusion according to claim 1, characterized in that: The particle filter algorithm in step 10 includes: Initialize the particle filter and randomly generate particles within the known initial position range; State prediction based on IMU data, applying the motion information generated by IMU data to each particle to simulate the possible movement path of the tag; Multimodal observation update of VLC and ultrasonic data, using RSSI, phase information of VLC signal and ultrasonic ranging value to update particle weights; Particle resampling: resample particles based on weights, retain high-weight particles, generate new particles near these particles, and abandon low-weight particles; Position output and dynamic update: The final position output obtained by weighted average calculation is continuously integrated with new data to achieve dynamic tracking of tag position.

5. An indoor positioning system based on dynamic mode switching and data fusion, used to execute the indoor positioning method based on dynamic mode switching and data fusion as described in any one of claims 1 to 4, characterized in that: include: Data acquisition module, including VLC signal acquisition submodule, IMU data acquisition submodule and ultrasonic ranging submodule; Data preprocessing module, including signal denoising submodule and feature extraction submodule; The signal quality monitoring and mode switching module includes a VLC signal detection submodule and a mode switching control submodule, which is used to monitor the VLC signal strength and stability in real time and select the positioning mode according to fuzzy logic rules; Data fusion module, including Kalman filter algorithm and particle filter algorithm, used to fuse VLC, IMU and ultrasonic data; Position estimation and output module, used to output the tag position information obtained after data fusion in real time; The system control module is used to manage the cyclic update operation of the system and control the execution order of modules such as data acquisition, preprocessing, and fusion.

6. The indoor positioning system based on dynamic mode switching and data fusion according to claim 5, characterized in that: The particle filter algorithm updates particle weights through multimodal observations and uses resampling technology to reduce positioning errors, thereby achieving high-precision positioning under long-term operation.

Images