Underground combined positioning method based on UWB and IMU

Abstract

The invention discloses an underground combined positioning method based on UWB and IMU, and the method comprises the following steps: arranging a UWB base station and a motion capture system in a positioning area, obtaining the distance between the base station and a label through UWB distance measurement, collecting IMU speed data through the movement of a person, recording the position of the label on the person through the motion capture system, and constructing a data set; constructing a positioning model based on the data set and the bidirectional long-short-term memory neural network; positioning the personnel through the positioning model to obtain a predicted positioning result; and fusing the predicted positioning result and IMU data through extended Kalman filtering to obtain a final positioning result. According to the invention, the accurate position of the underground coal mine personnel can be accurately positioned.

Description

Technical Field

[0001] The present invention belongs to the technical field of underground mine positioning, and in particular relates to an underground mine combined positioning method based on UWB and IMU. Background Art

[0002] As one of the "Six Major Systems," the underground coal mine personnel positioning system plays a vital role in coal mine production scheduling, safety management, and accident emergency rescue. Therefore, the ability to accurately monitor the location of underground coal mine personnel is of great significance to the development of the national coal industry and the construction and management of mine mining.

[0003] Ultra-wideband (UWB) positioning technology uses narrow pulse radio signals with extremely wide spectrum and very short time intervals for communication and ranging. Because the radio wave period used is in the nanosecond range, it has high temporal resolution, achieving centimeter-level distance resolution. UWB positioning systems perform well in complex indoor environments and are applicable not only to stationary objects but also to moving objects. They have broad prospects in mobile positioning, achieving centimeter-level accuracy when locating and tracking moving objects. However, in non-line-of-sight environments, UWB signals can experience significant offset or even loss, significantly reducing the accuracy and stability of UWB positioning technology.

[0004] An inertial navigation system (INS), based on classical Newtonian mechanics, utilizes inertial measurement units (IMUs), such as gyroscopes and accelerometers, to measure the linear and angular motion of a vehicle. Based on the vehicle's initial state, it then uses integral calculations to infer navigation information, including the vehicle's current position, velocity, and attitude. Once initialized, an INS is an autonomous navigation system that no longer relies on any external information and does not radiate energy. It can continuously provide various navigation parameters, such as the vehicle's position, velocity, and attitude, in any medium and environment worldwide. However, an INS also has a significant drawback: navigation errors accumulate over time. Therefore, an INS cannot be used alone for extended periods; it must be continuously calibrated to provide reliable navigation information. Summary of the Invention

[0005] The present invention proposes a combined positioning method for underground mines based on UWB and IMU to solve the problems existing in the above-mentioned prior art.

[0006] To achieve the above objectives, the present invention provides a combined positioning method for underground mines based on UWB and IMU, comprising the following steps:

[0007] Deploy UWB base stations and motion capture systems in the positioning area. Use UWB ranging to obtain the distance between the base station and the tag. Collect IMU velocity data from the movement of people. Use the motion capture system to record the location of the tag on the person and construct a data set.

[0008] Building a positioning model based on the data set and a bidirectional long short-term memory neural network;

[0009] Positioning the personnel using the positioning model to obtain a predicted positioning result;

[0010] The predicted positioning result is fused with the IMU data through extended Kalman filtering to obtain the final positioning result.

[0011] Preferably, when arranging UWB base stations, the tag can receive signals from more than 6 base stations, and the data set includes ranging data and real coordinate data; wherein, the ranging data represents the distance measurement values of the 6 UWB base stations and the tag at different time steps, and the real coordinate data represents the three-dimensional space coordinates of the tag at the last time step.

[0012] Preferably, the constructing of a positioning model based on the data set and the bidirectional long short-term memory neural network includes:

[0013] Constructing an input end and an output end of a bidirectional long short-term memory neural network according to the format of the data set;

[0014] The bidirectional long short-term memory neural network consists of two independent long short-term memory neural networks; the forward long short-term memory neural network is used to process the forward information of the input sequence; the reverse long short-term memory neural network is used to process the reverse information of the input sequence;

[0015] The bidirectional long short-term memory neural network controls the flow of information through input gates, forget gates, output gates and memory units.

[0016] Preferably, the positioning model further includes an attention mechanism and is connected to two fully connected layers with Dropout to correct the time series data error; for extracting,

[0017] The attention mechanism obtains key features by dynamically assigning different weights to each time step;

[0018] The final positioning result is obtained by calculating the key positioning features through the Chan algorithm.

[0019] Preferably, the loss function of the positioning model during training is:

[0020]

[0021] Where avg is the average, X, Y, Z are the real coordinates, X G ,Y G ,Z G are the coordinates calculated by the positioning model, λ is the regularization strength, and i,j represent the rows and columns of the weight matrix W.

[0022] Preferably, the fusion by extended Kalman filtering includes:

[0023] Obtain accurate motion information in a short period of time from IMU data and predict the state of motion;

[0024] The position coordinates output by the positioning model are used as observation values to correct the predicted motion state and obtain a fused high-precision positioning result.

[0025] The present invention also provides a combined positioning system for underground mines based on UWB and IMU, comprising:

[0026] The data acquisition module is used to deploy UWB base stations and motion capture systems in the positioning area, obtain the distance between the base station and the tag through UWB ranging, collect IMU speed data from the movement of people, record the location of the tag on the person through the motion capture system, and construct a data set;

[0027] A model building module, configured to build a positioning model based on the data set and a bidirectional long short-term memory neural network;

[0028] A prediction module, configured to locate a person using the positioning model and obtain a predicted positioning result;

[0029] The fusion module is used to fuse the predicted positioning result with the IMU data through extended Kalman filtering to obtain the final positioning result.

[0030] The present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0031] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method when executed by a processor.

[0032] The present invention also provides a computer program product, comprising a computer program, which implements the steps of the method when executed by a processor.

[0033] Compared with the prior art, the present invention has the following advantages and technical effects:

[0034] This invention achieves centimeter-level high-precision positioning by integrating ultra-wideband technology with inertial measurement unit data and employing an extended Kalman filter algorithm. This method not only enhances positioning stability and robustness, effectively reducing the impact of non-line-of-sight errors and multipath effects, but also pre-processes UWB data through a bidirectional long-short-term memory neural network and an attention mechanism, further improving the system's accuracy and adaptability to complex environments. Furthermore, this invention is cost-effective and provides strong support for mine safety management, dynamic personnel tracking, and emergency rescue, significantly improving the safety and efficiency of mine operations and advancing the development of intelligent mines. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0036] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0037] Figure 2 This is a diagram of an experimental scene according to an embodiment of the present invention;

[0038] Figure 3 A diagram illustrating the arrangement of a motion capture system according to an embodiment of the present invention;

[0039] Figure 4 This is a trajectory error analysis diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0040] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0041] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0042] Example 1

[0043] like Figure 1 As shown, this embodiment provides a combined positioning method for underground mines based on UWB and IMU, including the following steps:

[0044] Deploy UWB base stations and motion capture systems in the positioning area. Use UWB ranging to obtain the distance between the base station and the tag. Collect IMU velocity data from the movement of people. Use the motion capture system to record the location of the tag on the person and construct a data set.

[0045] Build a positioning model based on the dataset and bidirectional long short-term memory neural network;

[0046] Use the positioning model to locate the personnel and obtain the predicted positioning results;

[0047] The predicted positioning result is fused with the IMU data through the extended Kalman filter to obtain the final positioning result.

[0048] In this embodiment, a tunnel with a length of 30m, a width of 2m and a height of 2m is selected. Figure 2 shown.

[0049] The specific steps are as follows:

[0050] 1. Place UWB base stations according to the positioning system. Select a suitable base station as the primary base station in the experiment and connect it to the host computer to collect coordinate data. To verify the positioning accuracy of the algorithm in non-line-of-sight (NLOS) environments, a person will block the UWB device and walk around the experimental area during ranging. This way, the distance information obtained includes non-line-of-sight errors and is more consistent with actual application conditions.

[0051] 2. Establish a positioning model corresponding to the dataset

[0052] 3. The UWB data set is divided into a test set and a training set. The training set data is used to train the model. By continuously adjusting the parameters, the generalization of the model is determined in the test set. The dog model is then verified for predictive positioning.

[0053] 4. Combine the output of the positioning model with the IMU data and perform data fusion through the extended Kalman filter (EKF) to obtain the final positioning result.

[0054] Furthermore, step 1 specifically includes:

[0055] In the experimental site, choose the appropriate location to arrange the base station. What is needed is that there can be 6 UWB base stations at each location, which are respectively recorded as base stations A, B, C, D, E, and F. Use the motion capture system. The arrangement position of the motion capture system is as follows: Figure 3 As shown, the coordinates of each base station and the real-time location of personnel movement are captured. The format of the dataset is as follows:

[0056] (D 1 AG 、D 1 BG 、D 1 CG 、D 1 DG 、D 1EG 、D 1 FG ),...,(D 10 AG 、D 10 BG 、D 10 CG 、D 10 DG 、D 10 EG 、D 10 FG )|(X,Y,Z)(1)

[0057] Among them D i AG ,D i BG ,D i CG ,D i DG ,D i EG ,D i FG Respectively represent the distances from base stations A, B, C, D, E, and F to tag G at the i-th moment, where i ranges from 1 to 10. X, Y, and Z represent the three-dimensional coordinates of tag G at the last moment.

[0058] The tag coordinates measured by the motion capture system are used as the true values (data size is 1*3), and the coordinates of the tag and base station are obtained using the UWB device (data size is 6*10). Following this step, the dataset is expanded to 1000 and created as a dataset.

[0059] Furthermore, step 2 specifically includes:

[0060] First, define the input and output of the model. According to the style of the dataset, the size of the input is 6*10 and the size of the output is 1*3.

[0061] Next, we define the structure of a bidirectional long short-term memory (BiLSTM) neural network. BiLSTM consists of two independent LSTM networks: a forward LSTM that processes the forward information of the input sequence (from the past to the future). A reverse LSTM that processes the reverse information of the input sequence (from the future to the past). By combining the hidden states of the forward and reverse LSTMs, BiLSTM is able to capture the global context of the input sequence.

[0062] The bidirectional long short-term memory neural network mainly controls the flow of information by the following parts: input gate, forget gate, output gate, and memory unit.

[0063] Bidirectional long short-term memory neural network (BiLSTM) usually processes data through the following steps:

[0064] Step 1: Collect time series data and input the data into the bidirectional long short-term memory neural network. The input layer is expressed as follows: the ranging data of 6 base stations has a shape of 6×10 (6 base stations, 10 time steps)

[0065] x t =[x1,x2,...,x 10 ] (2)

[0066] Where Xt is the observation value or signal at different time t.

[0067] Step 2: Construct the input gate of the bidirectional long short-term memory neural network to extract and process the features of the input data. The calculation formula of the input gate is as follows:

[0068] y t =σ(Wf·[ht-1,xt]+bf) (3)

[0069] where y t represents the output of the input gate operation, and σ is the activation function.

[0070] Step 3: Construct the memory unit in the bidirectional long short-term memory neural network. The memory unit is the key to LSTM and ensures that the network can transmit information for a long time. The calculation formula is as follows:

[0071]

[0072] in, Represents the candidate memory cell state at the current time step; c t Represents the memory cell state at the current time step; W cx and W hc represents the weight matrix input in the memory cell state update; b c Represents the bias term in the memory cell state update. ⊙ represents the element-wise product, which represents the multiplication operation of the elements at the corresponding positions.

[0073] Step 4: Connect the data after residual connection using two fully connected layers. The function of the fully connected layer is to perform feature conversion and output prediction, thereby completing sequence modeling. The calculation formula of the fully connected layer is the same as that of the traditional fully connected layer, namely:

[0074]

[0075] a i =f(z i ) (7)

[0076] Where N is the number of neurons in the previous layer, ω ij is the weight connecting the jth input neuron and the ith output neuron. j is the output of the jth input neuron, b i is the bias of the i-th output neuron, z i is the weighted input, which represents the weighted sum of the inputs of the i-th neuron. f(·) is the activation function, which is usually a nonlinear function such as ReLu. i is the output of the i-th neuron, processed by the activation function.

[0077] Step 5: Output the predicted results.

[0078] By combining the hidden states of the forward and reverse LSTMs, the BiLSTM can capture the global context of the input sequence, better grasp the information between the data, and improve the accuracy of the model. The forward hidden state formula and the reverse hidden state formula are as follows:

[0079] Forward hidden state formula:

[0080] h t forward =LSTM(x t ,h t-1 f ) (8)

[0081] Reverse hidden state formula:

[0082] h t backward =LSTM(x t ,h t+1 b ) (9)

[0083] Among them, h t forward and h t backward They represent the hidden states of the temporal convolutional network in the forward and reverse directions respectively. The forward hidden state is mainly used to capture past information and sequence trends, while the reverse hidden state is used to capture future information and sequence trends.

[0084] An attention mechanism is also added after the network to help the network analyze the key features extracted by the BiLSTM layer, making the neural network's predictions more accurate. The attention mechanism consists of three important parts: the query (Q), which represents the current target item (the item for which attention weights need to be calculated); the key (K), which represents each item in the input sequence; and the value (V), which is the final output of the attention mechanism. Two fully connected layers with dropout are then connected to correct errors in the time series data to obtain the corrected data. The Chan algorithm solves a system of equations using a bilinear system of equations, but its performance degrades in the presence of high noise. BiLSTM is used to correct errors in the time series data, and the Chan algorithm is used to calculate the final positioning result.

[0085] Furthermore, step 3 specifically includes:

[0086] Define the loss function and select the Adam optimizer to train the model: Define a loss function that combines mean square error and regularization to train the deep learning model:

[0087]

[0088] Where avg is the average, X, Y, Z are the real coordinates, X G ,Y G ,Z G are the coordinates calculated by the positioning system, λ is the regularization strength, a non-negative hyperparameter, and i and j represent the rows and columns of the weight matrix W. The weighted mean squared error loss and the regularization loss are summed to obtain the final total loss. This total loss will serve as the optimizer's objective.

[0089] The core formula of the Adam optimizer is as follows:

[0090] 1. Initialize model parameters θ, learning rate α, momentum decay coefficient β1 (usually set to 0.9), β2 (usually set to 0.999).

[0091] 2. Then calculate the gradient:

[0092] g t =▽ θ J(θ t ) (11)

[0093] g t is the gradient of the current time step, θ t Current parameters, J(θ t ) is the loss function.

[0094] 3. Update the first-order moment estimate and the second-order moment estimate:

[0095]

[0096] m t is the first-order momentum of the current time step, v t is the second-order momentum at the current time step.

[0097] 4. Correction deviation:

[0098]

[0099] in is the corrected first-order moment estimate, is the corrected second moment estimate.

[0100] 5. Update model parameters:

[0101]

[0102] ξ is a small constant added for numerical stability. Repeat the above steps until the stopping condition is met.

[0103] Furthermore, step 4 specifically includes:

[0104] The output of the BiLSTM model is used as the observation value and the IMU data is used as the predicted value.

[0105] Prediction: Use IMU data to provide accurate motion information (such as acceleration and angular velocity) over a short period of time for predicting state.

[0106] Update: Use the output of the BiLSTM model to provide absolute position information for updating the state.

[0107] (1) Prediction steps:

[0108] Status prediction:

[0109]

[0110] The predicted state at the current moment, f nonlinear state transfer function, The state estimate at the previous moment, u k Control input (acceleration and angular velocity from the IMU).

[0111] Covariance prediction:

[0112]

[0113] The current moment's forecast covariance, F k State transfer matrix (Jacobian matrix of f), P k-1 The covariance matrix at the previous moment, Qk Process noise covariance matrix.

[0114] (2) Update steps:

[0115] Kalman gain calculation:

[0116]

[0117] K k Kalman gain, H k The measurement matrix (Jacobian matrix of h), R k Observation noise covariance matrix.

[0118] Status Update:

[0119]

[0120] The current state estimate, z k The observation value at the current moment (the position measured by UWB), h is the nonlinear observation function.

[0121] Covariance update:

[0122]

[0123] P k The covariance matrix at the current moment, I is the identity matrix.

[0124] The UWB and IMU data are fused through EKF to obtain the final positioning result.

[0125] 1000 points of the test data set were selected for experiment and experimental results were obtained. The results showed that the positioning error was within 7.1cm. Figure 4 The error comparison diagram between the actual trajectory and the calculated trajectory is shown.

[0126] This embodiment also provides a combined positioning system for underground mines based on UWB and IMU, including:

[0127] The data acquisition module is used to deploy UWB base stations and motion capture systems in the positioning area, obtain the distance between the base station and the tag through UWB ranging, collect IMU speed data from the movement of people, record the location of the tag on the person through the motion capture system, and construct a data set;

[0128] A model building module, used to build a positioning model based on the dataset and a bidirectional long short-term memory neural network;

[0129] The prediction module is used to locate people through the positioning model and obtain the predicted positioning results;

[0130] The fusion module is used to fuse the predicted positioning results with the IMU data through the extended Kalman filter to obtain the final positioning result.

[0131] This embodiment further provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0132] This embodiment further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method when executed by a processor.

[0133] This embodiment also provides a computer program product, including a computer program, which implements the steps of the method when executed by a processor.

[0134] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A combined positioning method for mines based on UWB and IMU, characterized in that: The following steps are involved: Deploy UWB base stations and motion capture systems in the positioning area. Use UWB ranging to obtain the distance between the base station and the tag. Collect IMU velocity data from the movement of people. Use the motion capture system to record the location of the tag on the person and construct a data set. Building a positioning model based on the data set and a bidirectional long short-term memory neural network; Positioning the personnel using the positioning model to obtain a predicted positioning result; The predicted positioning result is fused with the IMU data through extended Kalman filtering to obtain the final positioning result.

2. The method according to claim 1, characterized in that When UWB base stations are arranged so that the tag can receive signals from more than six base stations, the data set includes ranging data and real coordinate data; wherein the ranging data represents the distance measurement values between the six UWB base stations and the tag at different time steps, and the real coordinate data represents the three-dimensional spatial coordinates of the tag at the last time step.

3. The method according to claim 1, characterized in that The constructing of a positioning model based on the data set and the bidirectional long short-term memory neural network includes: Constructing an input end and an output end of a bidirectional long short-term memory neural network according to the format of the data set; The bidirectional long short-term memory neural network consists of two independent long short-term memory neural networks; the forward long short-term memory neural network is used to process the forward information of the input sequence; the reverse long short-term memory neural network is used to process the reverse information of the input sequence; The bidirectional long short-term memory neural network controls the flow of information through input gates, forget gates, output gates and memory units.

4. The method according to claim 3, characterized in that The positioning model also includes an attention mechanism and connects two fully connected layers with Dropout to correct the time series data error; it is used to extract, The attention mechanism obtains key features by dynamically assigning different weights to each time step; The final positioning result is obtained by calculating the key positioning features through the Chan algorithm.

5. The method according to claim 1, wherein The loss function of the positioning model during training is: Where avg is the average, X, Y, Z are the real coordinates, X G ,Y G ,Z G are the coordinates calculated by the positioning model, λ is the regularization strength, and i,j represent the rows and columns of the weight matrix W.

6. The method according to claim 1, characterized in that The fusion by extended Kalman filtering includes: Obtain accurate motion information in a short period of time from IMU data and predict the state of motion; The position coordinates output by the positioning model are used as observation values to correct the predicted motion state and obtain a fused high-precision positioning result.

7. A combined positioning system for mines based on UWB and IMU, characterized in that: include: The data acquisition module is used to deploy UWB base stations and motion capture systems in the positioning area, obtain the distance between the base station and the tag through UWB ranging, collect IMU speed data from the movement of people, record the location of the tag on the person through the motion capture system, and construct a data set; A model building module, configured to build a positioning model based on the data set and a bidirectional long short-term memory neural network; A prediction module, configured to locate a person using the positioning model and obtain a predicted positioning result; The fusion module is used to fuse the predicted positioning result with the IMU data through extended Kalman filtering to obtain the final positioning result.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images