Underground coal mine intelligent unmanned vehicle positioning method and system

Abstract

The invention discloses a coal mine underground intelligent unmanned vehicle positioning method and system, and belongs to the technical field of navigation positioning, and the method comprises the steps: constructing a neural network to assist a GNSS and INS combination system to position an unmanned vehicle in a ground area; in an underground working area, performing robust improvement on a UWB positioning algorithm to obtain a high robust UWB, and constructing a UWB and INS tight combination system of a high robust double-layer fusion architecture to position the unmanned vehicle; and a weighted fusion high-robust UWB, GNSS and INS combined system is constructed in a ground and underground transition area to position the unmanned vehicle. According to the invention, in consideration of complementarity of different sensors and internal and external positioning requirements of the mine, in different fields, different systems are arranged for positioning so as to realize smooth transition of positioning.

Description

Technical Field

[0001] The present invention belongs to the technical field of navigation and positioning, and relates to a positioning method and system for an intelligent unmanned vehicle in a coal mine. Background Art

[0002] In open outdoor environments, the Global Navigation Satellite System (GNSS) is sufficient for daily positioning needs. However, in coal mines, using a single sensor has significant limitations, making it difficult to achieve continuous navigation and positioning. This requires the coordinated operation of multiple sensors.

[0003] Invention patent CN 116734845 A proposes a seamless positioning method for mine vehicles based on a GNSS / UWB fusion IMU. This method uses GNSS fusion IMU positioning in the surface area, UWB fusion IMU positioning in the underground area, and confidence fusion positioning in the surface-underground interaction area. This method can achieve full-process positioning of mine vehicles in multiple spatiotemporal scenarios. However, this invention fails to consider that signal coverage is limited by interference from terrain and obstacles in the complex environment of a mine. Consequently, the error correction and correction mechanism may be insufficient, and error accumulation cannot be effectively suppressed.

[0004] Invention patent CN 111479233 B proposes a detection and position feedback system for unmanned mine vehicles based on UWB technology. This system collects environmental information and wirelessly communicates with a single-reference point positioning base station module within the area to obtain location information. Ultimately, this information is transmitted as a wireless signal. However, this system lacks the compensation capabilities of an inertial navigation system. Using a single UWB system, it is difficult to compensate for persistent errors during movement, and is prone to failure in environments with poor signal quality, severe occlusion, or significant multipath effects.

[0005] Invention patent CN 110645979 B proposes a seamless indoor and outdoor positioning method based on a GNSS / INS / UWB combination. This method determines whether the user is in an outdoor or indoor scene, and uses a combination of GNSS and INS for positioning in outdoor scenes, and a combination of UWB and INS for positioning in indoor scenes. When the UWB signal is valid indoors and the GNSS signal is invalid, the difference between the original distance measurement of the UWB and the distance calculated by the INS is used as the measurement input of the Kalman filter, and the output value of the Kalman filter is used to correct the solution result of the INS. However, this invention does not take into account the error accumulation of the INS system during long-term use, especially in long-term indoor positioning, the accumulated error of the INS will lead to a large positioning deviation.

[0006] Compared to conventional indoor environments, signal propagation in coal mines is severely attenuated by ore, rock formations, and metal scaffolding, making it difficult for high-frequency signals to penetrate underground structures. Furthermore, the numerous obstacles, complex tunnel structures, and equipment interference within mines cause severe multipath effects and signal reflections, significantly reducing positioning accuracy. Therefore, developing an improved fusion algorithm to address CNSS signal loss, UWB sensor delay deviation, and NLOS errors is crucial for improving system stability and positioning accuracy.

[0007] Therefore, the present invention proposes a method and system for positioning an intelligent unmanned vehicle in a coal mine. Summary of the Invention

[0008] The purpose of the present invention is to provide a method and system for positioning an intelligent unmanned vehicle in a coal mine, so as to overcome the shortcomings of the prior art in complex underground coal mine areas, such as poor stability and low positioning accuracy.

[0009] In order to achieve the above object, the present invention provides the following technical solutions: In a first aspect, the present invention provides a method for positioning an intelligent unmanned vehicle in a coal mine, which specifically comprises the following steps: In the surface area above the well, a neural network-assisted GNSS and INS combined system is constructed to locate the unmanned vehicle; In underground working areas, the UWB positioning algorithm is improved to obtain highly robust UWB and a highly robust dual-layer fusion architecture of UWB and INS is constructed to locate the unmanned vehicle. In the transition area between the surface and underground, a weighted fusion high-resistance UWB, GNSS and INS combined system is constructed to locate the unmanned vehicle.

[0010] Furthermore, in the above-ground area above the well, constructing a neural network-assisted GNSS and INS combined system to locate the unmanned vehicle specifically includes: S11, obtain the GNSS measurement value and INS output value; S12, establishing a GNSS and INS combined system model to uniformly describe GNSS measurements and INS outputs; the GNSS measurements include the position and velocity of the unmanned vehicle, and the INS outputs include the position, velocity, and attitude of the unmanned vehicle; S13: Establish a geometric relationship model between the INS and the satellite, convert the position and velocity output by the INS into the corresponding pseudorange and pseudorange rate, and make a difference between the pseudorange and pseudorange rate output by the GNSS. The difference is used as the observation value and input into the Kalman filter for correction. The GNSS and INS combined system completes the positioning; S14, if the GNSS signal is lost during positioning by the GNSS and INS combined system, a neural network is used to provide compensation information to replace the lost GNSS signal to assist in positioning.

[0011] Furthermore, the S14 specifically includes: If the GNSS signal is lost during positioning by the combined GNSS and INS system, the position, velocity, and attitude of the unmanned vehicle output by the INS are used as input to the neural network. During neural network training, the GNSS position increments at adjacent moments are used as target values ​​to design and optimize network parameters. The neural network uses LSTM network to process time series data. The attention mechanism layer is calculated after the LSTM layer and connected to the output layer. Calculate the attention weight on the hidden layer state of the LSTM network to indicate the degree of attention paid to historical information at the current moment. Multiply the attention weight by the hidden state of the practice step to obtain the hidden layer state at time T. , and then calculate the attention weight of each hidden layer state; The hidden layer state and the attention weight corresponding to each hidden layer state are used as input vectors. After being processed by the attention mechanism layer, the output vector is the weight vector, which represents the prediction of the GNSS position increment at the current moment. The weight vector is multiplied by the hidden layer state and then input into the fully connected layer to obtain the predicted value, which serves as the compensation information of the GNSS and INS combined system, thus completing the training of the neural network. In the GNSS failure area, the position, velocity and attitude calculated by INS are input into the trained neural network to predict the GNSS increment. The predicted GNSS increment value is then input into the Kalman filter for information fusion to assist the positioning of the GNSS and INS combined system.

[0012] Furthermore, the steps of improving the robustness of the UWB positioning algorithm are specifically as follows: S21, determine the position and speed of the unmanned vehicle, obtain the TOA observation value between the unmanned vehicle and the base station through the UWB sensor, initialize the position and speed of the unmanned vehicle, and generate the initialization covariance matrix; S22, predict the next position and speed of the unmanned vehicle based on its current position and speed, obtain the observation residual value and use it to initialize the covariance matrix, and update the Kalman gain expression and state update expression at the same time; S23, judge whether the observed residual value is obvious, if not obvious, skip it, if obvious, use IGG III method to update the weight; S24, calculate the Kalman gain matrix by using the IGG III function as the equivalent weight function in the robust EKF process, and introduce the variance inflation factor G in the calculation; S25, determining the abnormality level of the UWB system, wherein the abnormality level includes general abnormality observation and system-level abnormality observation; S26, when the observation value of the UWB system shows system-level abnormal observation, the covariance matrix is ​​re-evaluated through the innovation vector and the robust iterative calculation is stopped; S27, when the observation value of the UWB system shows a general abnormal observation, the residual value of each observation value is calculated and recorded, and the predicted observation value is obtained according to the current state of the unmanned vehicle. Based on the calculated residual value, the state estimate is corrected through the standard EKF measurement update step, and the covariance matrix is ​​updated; S28, the UWB system determines whether it has converged based on the n-iteration estimation value and the n+1-iteration estimation value corresponding to the current observation value. If it has not converged, it returns to S23 to iterate until it converges. If it has converged, it ends the iteration and completes the high-resistance improvement of UWB.

[0013] Furthermore, the improved highly robust UWB is used to construct a highly robust dual-layer fusion architecture of UWB and INS tightly combined system for positioning the unmanned vehicle, specifically including: The improved high-resistance UWB is used to construct a highly robust dual-layer fusion architecture of UWB and INS tightly combined system for positioning the unmanned vehicle, specifically including: Construct state variables and obtain state variable differential equations; Discretize the state variable differential equation to obtain the discrete time system state equation; The center position of the highly robust UWB tag antenna is obtained through the INS navigation results and the arm measurement value; Calculate the Euclidean straight-line distance between the tag antenna center and the location of the high-resistance UWB base station; Obtaining observation information, the observation information including the navigation result of the INS and the Euclidean straight-line distance between the center of the tag antenna and the position of the high-robust UWB base station; The measurement equation is obtained based on the discrete time system state equation, the center position of the high-robust UWB tag antenna, and the Euclidean straight-line distance between the tag antenna center and the high-robust UWB base station position; Add the observation information to the low-level architecture for low-level fusion to obtain the preliminary positioning results after low-level fusion; The preliminary positioning results are passed to the high-level architecture, and the environmental perception information is added to the high-level architecture for high-level fusion; Environmental perception information senses the terrain and adjusts the fusion weight of the highly robust UWB and INS, and re-integrates the signals based on the adjusted weight information to give the positioning result.

[0014] Furthermore, the specific process of adding the observation information to the low-level architecture for low-level fusion and obtaining the preliminary positioning result after low-level fusion is as follows: On the basis of the measurement equation, a multi-factor filtering algorithm is introduced to fuse the observation information; Calculate the residual of the observation information and construct the weight coefficient according to the size of the residual; Test the degree of deviation between the observed information and the actual measured information. The specific test formula is: ; in, is the innovation vector of the compact combination system, is the covariance matrix of the system innovation vector; The inverse of the error covariance matrix of each observation information is calculated, and the normalized inverse of the error covariance matrix of all observation information is calculated to construct an equivalent weight factor, which is used to dynamically adjust the weight of each observation information to obtain the preliminary positioning result after low-level fusion.

[0015] Furthermore, the formula for adjusting the fusion weight of high-robust UWB and INS is: ; ; ; in, is the weight adjustment coefficient, is the actual slope value of the current terrain, is the slope threshold, is the preset maximum slope value. is the fusion weight of INS, is the fusion weight of UWB.

[0016] Furthermore, in the transition area between the surface and underground, the unmanned vehicle is positioned by using a weighted fusion high-resistance UWB, GNSS and INS combined system, specifically including: S31, respectively calculate the weights of GNSS, highly robust UWB and INS at the current moment; S32, calculate the comprehensive weight of the three by weighted average method; S33 dynamically adjusts the corresponding weights of GNSS, high-resistance UWB, and INS according to their quality indicators to ensure the final positioning result in the transition area between the wellbore and the wellbore.

[0017] Furthermore, the specific calculation process of S31 is: Get GNSS signal quality indicators from GNSS receivers and number of satellites , the comprehensive weight of GNSS is calculated by weighted average signal-to-noise ratio and number of satellites; ; in, is the GNSS comprehensive weight, is the GNSS signal quality indicator, is the number of satellites; Constructing multipath effect evaluation value To dynamically adjust the weight of the high-resistance UWB, the signal strength of the high-resistance UWB is , then the weight calculation formula of high robustness UWB is as follows: ; in, is the UWB comprehensive weight, is the UWB signal strength, is the multipath effect evaluation value; Through accelerometer noise and gyroscope noise Calculate the weight of INS , as shown in the following formula: ; in, is the comprehensive weight of INS, is the accelerometer noise, Gyroscope noise.

[0018] In a second aspect, the present invention provides an intelligent unmanned vehicle positioning system in a coal mine, comprising: Inoue positioning module: Build a neural network to assist the GNSS and INS combined system to locate the unmanned vehicle; Downhole positioning module: Improve the robustness of the UWB positioning algorithm to obtain highly robust UWB and build a highly robust dual-layer fusion architecture of UWB and INS tightly combined system to locate the unmanned vehicle; Transition area positioning module: Build a weighted fusion high-resistance UWB, GNSS and INS combined system to locate the unmanned vehicle.

[0019] Compared with the prior art, the present invention has the following beneficial technical effects: The present invention provides a method for positioning an intelligent unmanned vehicle in a coal mine. Taking into account the complementarity of different sensors and the needs of positioning inside and outside the mine, different systems are set up for positioning in different fields to achieve a smooth transition of positioning. In the ground area above the mine, a neural network is used to assist the GNSS and INS combined system to achieve positioning of the unmanned vehicle, thereby improving the robustness of positioning; in the underground working area, a UWB and INS tight combination system with a high-resistance dual-layer fusion architecture is used to position the unmanned vehicle; in the transition area between the mine and underground, a weighted fusion high-resistance UWB, GNSS and INS combined system is constructed to position the unmanned vehicle.

[0020] Specifically, a neural network is used to assist the GNSS and INS combined system, using an LSTM network to compensate for the signal output when the GNSS signal loses lock. The UWB and INS tight combination system adopts a high-resistance dual-layer fusion architecture. The low-layer fusion focuses on processing basic sensor data, and the high-layer fusion makes more precise adjustments and optimizations according to environmental changes. Layered processing can effectively isolate and control the influence of different error sources, thereby improving positioning accuracy. The high-resistance UWB, GNSS and INS combined system adjusts the fusion weights of each signal in real time according to the reception status of the GNSS signal and the UWB signal, achieving a smooth transition of positioning and ensuring the continuity of positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of a method for positioning an intelligent unmanned vehicle in a coal mine according to an embodiment of the present invention.

[0022] Figure 2 Schematic diagram of a neural network-assisted GNSS and INS combined system positioning method according to an embodiment of the present invention.

[0023] Figure 3 Schematic diagram of the improved high-resistance UWB positioning algorithm in an embodiment of the present invention.

[0024] Figure 4 Schematic diagram of a tightly combined UWB and INS system with a highly robust dual-layer fusion architecture in an embodiment of the present invention.

[0025] Figure 5 Schematic diagram of a highly robust UWB, GNSS, and INS combined system in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0027] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] like Figure 1 As shown, the present invention provides a method for positioning an intelligent unmanned vehicle in a coal mine, comprising the following steps: In the surface area above the well, a neural network-assisted GNSS and INS combined system is constructed to locate the unmanned vehicle; the neural network-assisted GNSS and INS combined system can take into account the situation where the GNSS signal is lost, such as Figure 2 As shown in the figure, in order to achieve the purpose of using neural network to assist GNSS / INS integrated navigation and positioning in the surface area above the well, the specific steps include: S11, obtain the GNSS measurement value and INS output value; The position, velocity, and time of the UAV are obtained from the raw satellite observation data provided by GNSS. The acceleration and angular velocity of the UAV are measured by the INS accelerometer and gyroscope to obtain the position, velocity, and attitude of the UAV. S12, establish a GNSS and INS combined system model to unify the description of GNSS measurements and INS outputs; GNSS measurements include the position and velocity of the unmanned vehicle, and INS outputs include the position, velocity, and attitude of the unmanned vehicle; The GNSS and INS combined system model is established using the state space equation, as shown in the following formula: ; in, is the state vector of INS, is the state vector of GNSS, is the state transition matrix of INS, is the state transition matrix of GNSS, is the noise driving matrix of INS, is the noise driving matrix of GNSS, is the process noise vector of the GNSS / INS integrated navigation system, is the time derivative of the INS state vector, is the time derivative of the GNSS state vector; S13, establish a geometric relationship model between INS and satellite, convert the position and velocity output by INS into corresponding pseudorange and pseudorange rate, and make a difference with the pseudorange and pseudorange rate output by GNSS; The pseudorange and pseudorange rate between INS and satellite are expressed as follows: ; ; in, is the pseudorange to the mth satellite calculated by INS, 、 、 The three-dimensional position coordinates of the unmanned vehicle in the Earth-centered Earth-fixed coordinate system output by INS, 、 、 is the three-dimensional position coordinate of the mth satellite in the Earth-centered Earth-fixed coordinate system, is the pseudorange rate to the mth satellite calculated by INS, 、 、 is the component of the unit line of sight vector of the INS pointing to the mth satellite in the Earth-centered Earth-fixed coordinate system, 、 、 is the three-dimensional velocity component of the unmanned vehicle in the Earth-centered Earth-fixed coordinate system output by INS, 、 、 is the three-dimensional velocity component of the mth satellite in the Earth-centered Earth-fixed coordinate system; The position and velocity output by the INS are converted into corresponding pseudoranges and pseudorange rates, and then subtracted from the pseudoranges and pseudorange rates output by the GNSS to obtain the difference. The difference is then input into the Kalman filter as an observation to perform corrections and optimize the estimation results of the unmanned vehicle's positioning, velocity, and attitude. The observation is shown in the following formula: ; Among them, Z is the observed quantity, is the pseudorange residual, is the pseudorange rate residual, 、 is the pseudorange of INS and GPS, 、 is the pseudorange rate of INS and GPS, G is the line-of-sight unit vector from the satellite to the receiver position output by the mechanical orchestration, T is the transformation from the geodetic coordinate system to the Cartesian coordinate system in the Earth-centered Earth-fixed coordinate system, R is the velocity conversion between the local horizontal coordinate system and the Earth coordinate system, and X is the state vector; S14, if the GNSS signal is lost during positioning by the GNSS and INS combined system, a neural network is used to provide compensation information to replace the lost GNSS signal to assist in positioning.

[0029] In some preferred embodiments of the present invention, the process of using a neural network to provide compensation information to replace the lost GNSS signal assisted positioning is specifically as follows: If the GNSS signal is lost during positioning by the combined GNSS and INS system, the position, velocity, and attitude of the unmanned vehicle output by the INS are used as input to the neural network. During neural network training, the GNSS position increments at adjacent moments are used as target values ​​to design and optimize network parameters: The neural network uses an LSTM network to process time series data to capture temporal dependencies and use them for predictions at future moments. An attention mechanism layer is calculated after the LSTM layer so that the neural network can automatically learn the importance of information between different time steps. The attention mechanism layer is connected to the output layer. Calculate the attention weight on the hidden layer state of the LSTM network to indicate the degree of attention paid to historical information at the current moment. Multiply the attention weight by the hidden state of the practice step to obtain the hidden layer state at time T. , and then calculate the attention weight of each hidden layer state; The attention weight is expressed as follows: ; ; in, Score for attention, is the weight vector, 、U is for general 、 The weight matrix mapped to the attention space, is the hidden state of the decoder at time k, is the temporal feature extracted at time i, b is the bias vector, C is the context vector, is the attention weight of the current moment k to the historical moment i; The hidden layer state and the attention weight corresponding to each hidden layer state are used as input vectors. After being processed by the attention mechanism layer, the output vector is the weight vector, which represents the prediction of the GNSS position increment at the current moment. The weight vector is expressed by the following formula: ; in, is the hidden layer state, is the state update function, is the context vector, Input features for the current moment; The weight vector is multiplied by the hidden layer state and then input into the fully connected layer to obtain the predicted value, which serves as the compensation information of the GNSS and INS combined system, thus completing the training of the neural network. In the GNSS failure area, the position, velocity and attitude calculated by the INS are input into the trained neural network to predict the GNSS increment. The predicted GNSS increment value is then input into the Kalman filter for information fusion. The GNSS and INS combined system enters the pure inertial navigation mode to prevent the positioning accuracy from decreasing rapidly, thereby assisting the GNSS and INS combined system in positioning.

[0030] like Figures 3 and 4 As shown in the figure, in the underground working area, the UWB positioning algorithm is improved to obtain a highly robust UWB and a highly robust dual-layer fusion architecture of UWB and INS is constructed to locate the unmanned vehicle. In order to address the error characteristics of different types of abnormal observations such as system errors and environmental errors, an improved UWB robust positioning algorithm is proposed. The specific steps include: S21, determine the position and speed of the unmanned vehicle, input the known coordinates of the base station into the observation system, obtain the TOA observation value between the unmanned vehicle and the base station through the UWB sensor, further initialize the position and speed of the unmanned vehicle, and generate an initialization covariance matrix; S22, predict the next position and speed of the unmanned vehicle based on its current position and speed, obtain the observation residual value and use it to initialize the covariance matrix, and update the Kalman gain expression and state update expression at the same time; S23, judge whether the observed residual value is obvious. If the residual value is close to zero, skip it. If it is obvious, use the IGG III method to update the weight. ; Original weight Initialized to 1; S24, calculate the Kalman gain matrix using the IGG III function as the equivalent weight function in the robust EKF process, and introduce the variance inflation factor G during the calculation to attenuate the weight of abnormal observations; In some preferred embodiments of the present invention, the calculation process of the Kalman gain matrix is ​​further adjusted to minimize the impact of abnormal observations on state estimation, and the corresponding covariance is further increased by the variance inflation factor, thereby effectively reducing the contribution of the observation to the Kalman gain matrix. In this case, the calculation process of the Kalman gain matrix is ​​shown in the following formula: ; in, is the Kalman gain matrix, is the forecast error covariance matrix, is the observation matrix, is the variance inflation factor matrix, is the observation noise covariance matrix; S25, determining the abnormality level of the UWB system, wherein the abnormality level includes general abnormality observation and system-level abnormality observation; S26: When a system-level abnormal observation occurs in the observation value of the UWB system, the covariance matrix is ​​re-evaluated through the innovation vector and adjusted according to the impact of the abnormal observation. The robust iterative calculation is stopped to prevent the iterative correction from causing erroneous state estimation. The covariance matrix is ​​adaptively adjusted according to the actual innovation information. By attenuating or weighting the abnormal observation, the impact of the abnormal data on the filtering result is minimized. S27, when the observation value of the UWB system has a general abnormal observation, the residual value of each observation value is calculated and recorded, and the predicted observation value is obtained according to the current state of the unmanned vehicle. Based on the calculated residual value, the state estimation is corrected through the standard EKF measurement update step, and the covariance matrix is ​​updated. The purpose of recording the residual value is to identify and process the abnormal observation in the future, especially when the residual value exceeds a certain threshold, it can be further determined whether the observation is an abnormal observation, and the corresponding anti-error processing is performed in the subsequent steps. It is necessary to calculate the EKF measurement update and record the residual value .

[0031] S28, the UWB system determines whether it has converged based on the n-iteration estimation value and the n+1-iteration estimation value corresponding to the current observation value. If the difference between the two iterative estimation values ​​is less than the set threshold, it means that the state estimation of the Kalman filter has stabilized, indicating that the observation system has converged, then the iteration ends and the high-resistance improvement of UWB is completed. If it has not converged, it returns to S23 to iterate until convergence.

[0032] Usually, convergence judgment is completed by setting a small tolerance value. For example, when the n-iteration estimation value and the n+1-iteration estimation value corresponding to the current observation value are less than a certain threshold, it is considered that convergence has occurred and the state estimation of the Kalman filter enters a stable state. Otherwise, weight update and state prediction correction are continued to improve the estimation accuracy.

[0033] Based on the improved high-resistance UWB, a highly robust dual-layer fusion architecture of UWB and INS is constructed to locate the unmanned vehicle. Specifically, the following are performed: Construct state variables and obtain state variable differential equations; By defining the zero bias error and time delay deviation of the inertial sensor as error terms, the error characteristics of the UWB and INS combined positioning system are fully described. The error terms are integrated into a 16-dimensional state vector and the state vector is updated and corrected, so that the deviations caused by various errors can be corrected in real time during each state estimation, thereby ensuring that the system can provide accurate positioning results and providing important data support for subsequent positioning optimization and error correction.

[0034] Obtaining a state variable differential equation based on a state vector, wherein the state vector includes the position, velocity, attitude, zero bias error of the inertial sensor, and time delay deviation of the inertial sensor of the unmanned vehicle; The state variables are shown in the following formula: ; in, is the position error, is the speed error, is the attitude angle error, is the gyroscope bias error, is the accelerometer bias error, is the receiver clock error; The state variable differential equation is: ; in, is the state transition matrix, is the error state vector, Noise driven matrix, is the system process noise vector; The state variable differential equation is discretized for real-time processing in a digital computing environment, thereby obtaining a continuous-time state variable error equation. Based on the continuous-time state variable error equation, a discrete-time system state equation is constructed. The discrete-time system state equation is: ; in, is the state error vector at the kth moment, is the discrete state transfer matrix, is the state error vector at the k-1th moment, is the discrete process noise vector; The center position of the highly robust UWB tag antenna is obtained through the INS navigation results and the arm measurement value; The specific process is as follows: Based on the navigation results provided by the INS, the position vector of the unmanned vehicle is obtained, which represents the three-dimensional position calculated by the inertial navigation system. The relative position from the UWB base station to the UWB tag or the length of the arm is obtained from the measurement arm. The arm measurement value is converted using a rotation matrix from the target coordinate system to the global coordinate system. The rotation matrix is ​​determined by the attitude information provided by the INS (such as roll angle, pitch angle, and yaw angle). Based on the INS navigation results and the arm measurement value, the center position of the UWB tag antenna is derived as follows:

[0035] in, Estimated position of the UWB tag antenna center, is the position vector of the unmanned vehicle in the navigation coordinate system output by INS, is the coordinate system rotation matrix, is the arm vector in the carrier coordinate system, is the velocity vector of the unmanned vehicle in the navigation coordinate system output by INS, is the time synchronization error; Calculate the Euclidean straight-line distance between the tag antenna center and the location of the high-resistance UWB base station; The details are shown in the following formula: ; in, is the Euclidean straight-line distance from the tag antenna center to the i-th UWB base station, 、 、 is the three-dimensional coordinate of the tag antenna center in the navigation coordinate system, 、 、 is the known three-dimensional coordinate of the i-th UWB base station in the same navigation coordinate system; Obtaining observation information, the observation information including the navigation result of the INS and the Euclidean straight-line distance between the center of the tag antenna and the position of the high-robust UWB base station; The measurement equation is obtained based on the discrete time system state equation, the center position of the high-robust UWB tag antenna, and the Euclidean straight-line distance between the tag antenna center and the high-robust UWB base station position; The measurement equation is: ; in, is the observation vector at the kth moment, is the observation matrix, is the error state vector at the kth moment, is the observation noise vector; Add the observation information to the low-level architecture for low-level fusion to obtain the preliminary positioning results after low-level fusion; The preliminary positioning results are passed to the high-level architecture, and the environmental perception information is added to the high-level architecture for high-level fusion; Environmental perception information senses the terrain and adjusts the fusion weight of the highly robust UWB and INS, and re-integrates the signals based on the adjusted weight information to give the positioning result.

[0036] In some preferred embodiments of the present invention, The observation information is added to the low-level architecture for low-level fusion, and the initial positioning result after low-level fusion is obtained. The specific process is as follows: Based on the measurement equation, a multi-factor filtering algorithm is introduced to fuse the observation information; the observation information is UWB ranging and INS navigation data; Calculate the residual of the observation information and construct the weight coefficient according to the size of the residual so that the observation information with small residual dominates the positioning result during fusion; Test the degree of deviation between the observed information and the actual measured information. The specific test formula is: ; in, is the residual of the ith observation, is the standardized residual, is the covariance matrix of the observed residuals; In order to quantitatively test the mathematical expression of observation information, the inverse of the error covariance matrix of each observation information is calculated to reflect the reliability of each observation information; a smaller covariance matrix means that the observation information is more reliable, so it is given a larger weight; the normalization of the inverse of the error covariance matrix of all observation information is calculated to construct an equivalent weight factor, which is used to dynamically adjust the weight of each observation information to obtain the preliminary positioning result after low-level fusion; construct an equivalent weight factor As shown in the following formula: ; in, Standardized residuals, is the maximum residual threshold, is the minimum residual threshold; After deep fusion of environmental perception information is introduced into the high-level architecture, the terrain slope or height difference is calculated. When the slope is greater than the threshold of 10%, it indicates that the terrain is rugged. The terrain slope and height difference are calculated as shown in the following formula: ; in, is the terrain slope, is the elevation value of the i-th terrain point, is the elevation value of the j-th terrain point, is the horizontal distance between the i-th point and the j-th point; When environmental perception detects rugged terrain, the fusion weight of UWB and INS is adjusted, taking into account that INS accuracy is significantly affected by elevation changes. The formula for adjusting the fusion weight of high-robust UWB and INS is: ; ; ; in, is the adjustment factor, is the terrain slope, is the slope threshold, is the maximum slope, is the INS fusion weight, is the UWB fusion weight.

[0037] In the transition zone between the surface and underground, the unmanned vehicle is positioned using a weighted fusion of a highly robust UWB, GNSS, and INS combined system. Specifically, the system includes: S31, respectively calculate the weights of GNSS, highly robust UWB and INS at the current moment; S32, calculates the combined weights of the three using a weighted average method. These weights are dynamically adjusted based on the quality indicators of each sensor to ensure that the sensor's impact on the final positioning result is consistent with its signal quality in the transition area between above and below the coal mine. S33 dynamically adjusts the corresponding weights of GNSS, high-resistance UWB and INS according to their quality indicators to ensure the final positioning result in the transition area between the wellbore and the wellbore; the final positioning result The result is shown in the following formula: ; ; in, is the sum of weights, 、 、 is the dynamic fusion weight of each sensor, 、 、 Independent positioning results for each sensor, The final fusion positioning result; As the environment changes, sensor signal quality fluctuates. By continuously acquiring data from each sensor and performing a weighted average based on real-time weights, the system calculates a fused positioning result at each moment. This process ensures that the system can flexibly adjust the contribution of each sensor based on current environmental changes.

[0038] In some preferred embodiments of the present invention, the specific calculation process of S31 is: Get GNSS signal quality indicators from GNSS receivers and number of satellites , the comprehensive weight of GNSS is calculated by weighted average signal-to-noise ratio and number of satellites; ; in, is the dynamic fusion weight of GNSS, is the GNSS signal quality indicator, is the number of satellites; Constructing multipath effect evaluation value To dynamically adjust the weight of the high-resistance UWB, the signal strength of the high-resistance UWB is , then the weight calculation formula of high robustness UWB is as follows: ; in, is the dynamic fusion weight of UWB, is the UWB signal strength, is the multipath effect evaluation value; Through accelerometer noise and gyroscope noise Calculate the weight of INS , as shown in the following formula: ; in, is the dynamic fusion weight of INS, is the accelerometer noise, is the gyroscope noise.

[0039] In a second aspect, the present invention further provides an intelligent unmanned vehicle positioning system for a coal mine, comprising: Inoue positioning module: Build a neural network to assist the GNSS and INS combined system to locate the unmanned vehicle; Downhole positioning module: Improve the robustness of the UWB positioning algorithm to obtain highly robust UWB and build a highly robust dual-layer fusion architecture of UWB and INS tightly combined system to locate the unmanned vehicle; Transition area positioning module: Build a weighted fusion high-resistance UWB, GNSS and INS combined system to locate the unmanned vehicle, and assign different weights to the sensors according to their signal strength and accuracy.

[0040] When the GNSS signal is good but the UWB signal is not, it is judged that the intelligent unmanned vehicle is in the ground area above the well, and the GNSS and INS combined system is used for positioning; when there is no GNSS signal and no UWB signal, it is judged that the intelligent unmanned vehicle is in the ground area above the well and the GNSS signal is locked, and the neural network / INS positioning is used; when there is a GNSS signal and the UWB signal is present, it is judged that the intelligent unmanned vehicle is in the transition area between the well and underground, and the weighted fusion high-resistance UWB, GNSS and INS combined system is used to position the unmanned vehicle; when there is no GNSS signal and the UWB signal is good, it is judged that the intelligent unmanned vehicle has entered the underground working area, and the UWB and INS tight combination system with a high-resistance dual-layer fusion architecture is used to position the unmanned vehicle, achieving high-precision and seamless positioning of the system.

[0041] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for positioning an intelligent unmanned vehicle in a coal mine, characterized in that: The following steps are involved: In the surface area above the well, a neural network-assisted GNSS and INS combined system is constructed to locate the unmanned vehicle; In underground working areas, the UWB positioning algorithm is improved to obtain highly robust UWB and a highly robust dual-layer fusion architecture of UWB and INS is constructed to locate the unmanned vehicle. In the transition area between the surface and underground, a weighted fusion high-resistance UWB, GNSS and INS combined system is constructed to locate the unmanned vehicle.

2. The method for positioning an intelligent unmanned vehicle in a coal mine according to claim 1, characterized in that: The construction of a neural network-assisted GNSS and INS combined system for positioning the unmanned vehicle in the surface area above the well specifically includes: S11, obtaining the GNSS measurement value and the INS output value; S12, establishing a GNSS and INS combined system model to uniformly describe GNSS measurements and INS outputs; the GNSS measurements include the position and velocity of the unmanned vehicle, and the INS outputs include the position, velocity, and attitude of the unmanned vehicle; S13: Establish a geometric relationship model between the INS and the satellite, convert the position and velocity output by the INS into the corresponding pseudorange and pseudorange rate, and make a difference between the pseudorange and pseudorange rate output by the GNSS. The difference is used as the observation value and input into the Kalman filter for correction. The GNSS and INS combined system completes the positioning; S14, if the GNSS signal is lost during positioning by the GNSS and INS combined system, a neural network is used to provide compensation information to replace the lost GNSS signal to assist in positioning.

3. The method for positioning an intelligent unmanned vehicle in a coal mine according to claim 2, characterized in that: The S14 specifically includes: If the GNSS signal is lost during positioning by the combined GNSS and INS system, the position, velocity, and attitude of the unmanned vehicle output by the INS are used as input to the neural network. During neural network training, the GNSS position increments at adjacent moments are used as target values ​​to design and optimize network parameters. The neural network uses an LSTM network to process time series data. The attention mechanism layer is calculated after the LSTM layer and connected to the output layer. Calculate the attention weight on the hidden layer state of the LSTM network to indicate the degree of attention paid to historical information at the current moment. Multiply the attention weight by the hidden state of the practice step to obtain the hidden layer state at time T. , and then calculate the attention weight of each hidden layer state; The hidden layer state and the attention weight corresponding to each hidden layer state are used as input vectors. After being processed by the attention mechanism layer, the output vector is the weight vector, which represents the prediction of the GNSS position increment at the current moment. The weight vector is multiplied by the hidden layer state and then input into the fully connected layer to obtain the predicted value, which serves as the compensation information of the GNSS and INS combined system, thus completing the training of the neural network. In the GNSS failure area, the position, velocity and attitude calculated by INS are input into the trained neural network to predict the GNSS increment. The predicted GNSS increment value is then input into the Kalman filter for information fusion to assist the positioning of the GNSS and INS combined system.

4. The method for positioning an intelligent unmanned vehicle in a coal mine according to claim 1, characterized in that: The steps for improving the robustness of the UWB positioning algorithm are specifically as follows: S21, determine the position and speed of the unmanned vehicle, obtain the TOA observation value between the unmanned vehicle and the base station through the UWB sensor, initialize the position and speed of the unmanned vehicle, and generate the initialization covariance matrix; S22, predict the next position and speed of the unmanned vehicle based on its current position and speed, obtain the observation residual value and use it to initialize the covariance matrix, and update the Kalman gain expression and state update expression at the same time; S23, judge whether the observed residual value is obvious, if not obvious, skip it, if obvious, use IGG III method to update the weight; S24, calculate the Kalman gain matrix by using the IGG III function as the equivalent weight function in the robust EKF process, and introduce the variance inflation factor G in the calculation; S25, determining the abnormality level of the UWB system, wherein the abnormality level includes general abnormality observation and system-level abnormality observation; S26, when the observation value of the UWB system shows system-level abnormal observation, the covariance matrix is ​​re-evaluated through the innovation vector and the robust iterative calculation is stopped; S27, when the observation value of the UWB system shows a general abnormal observation, the residual value of each observation value is calculated and recorded, and the predicted observation value is obtained according to the current state of the unmanned vehicle. Based on the calculated residual value, the state estimate is corrected through the standard EKF measurement update step, and the covariance matrix is ​​updated; S28, the UWB system determines whether it has converged based on the n-iteration estimation value and the n+1-iteration estimation value corresponding to the current observation value. If it has not converged, it returns to S23 to iterate until it converges. If it has converged, it ends the iteration and completes the high-resistance improvement of UWB.

5. The method for positioning an intelligent unmanned vehicle in a coal mine according to claim 1, characterized in that: The improved high-resistance UWB is used to construct a highly robust dual-layer fusion architecture of UWB and INS tightly combined system for positioning the unmanned vehicle, specifically including: Construct state variables and obtain state variable differential equations; Discretize the state variable differential equation to obtain the discrete time system state equation; The center position of the highly robust UWB tag antenna is obtained through the INS navigation results and the arm measurement value; Calculate the Euclidean straight-line distance between the tag antenna center and the location of the high-resistance UWB base station; Obtaining observation information, the observation information including the navigation result of the INS and the Euclidean straight-line distance between the center of the tag antenna and the position of the high-robust UWB base station; The measurement equation is obtained based on the discrete time system state equation, the center position of the high-robust UWB tag antenna, and the Euclidean straight-line distance between the tag antenna center and the high-robust UWB base station position; Add the observation information to the low-level architecture for low-level fusion to obtain the preliminary positioning results after low-level fusion; The preliminary positioning results are passed to the high-level architecture, and the environmental perception information is added to the high-level architecture for high-level fusion; Environmental perception information senses the terrain and adjusts the fusion weight of the highly robust UWB and INS, and re-integrates the signals based on the adjusted weight information to give the positioning result.

6. The method for positioning an intelligent unmanned vehicle in a coal mine according to claim 5, characterized in that: The specific process of adding observation information to the low-level architecture for low-level fusion and obtaining the preliminary positioning result after low-level fusion is as follows: On the basis of the measurement equation, a multi-factor filtering algorithm is introduced to fuse the observation information; Calculate the residual of the observation information and construct the weight coefficient according to the size of the residual; Test the degree of deviation between the observed information and the actual measured information. The specific test formula is: ； in, is the innovation vector of the compact combination system, is the covariance matrix of the system innovation vector; The inverse of the error covariance matrix of each observation information is calculated, and the normalized inverse of the error covariance matrix of all observation information is calculated to construct an equivalent weight factor, which is used to dynamically adjust the weight of each observation information to obtain the preliminary positioning result after low-level fusion.

7. The method for positioning an intelligent unmanned vehicle in a coal mine according to claim 5, characterized in that: The formula for adjusting the fusion weight of high-robust UWB and INS is: ； ； ； in, is the weight adjustment coefficient, is the actual slope value of the current terrain, is the slope threshold, is the preset maximum slope value. is the fusion weight of INS, is the fusion weight of UWB.

8. The method for positioning an intelligent unmanned vehicle in a coal mine according to claim 1, characterized in that: In the transition area between the surface and underground, the unmanned vehicle is positioned by weighted fusion of a highly robust UWB, GNSS, and INS combined system, specifically including: S31, respectively calculate the weights of GNSS, highly robust UWB and INS at the current moment; S32, calculate the comprehensive weight of the three by weighted average method; S33 dynamically adjusts the corresponding weights of GNSS, high-resistance UWB, and INS according to their quality indicators to ensure the final positioning result in the transition area between the wellbore and the wellbore.

9. The method for positioning an intelligent unmanned vehicle in a coal mine according to claim 1, characterized in that: The specific calculation process of S31 is: Get GNSS signal quality indicators from GNSS receivers and number of satellites , the comprehensive weight of GNSS is calculated by weighted average signal-to-noise ratio and number of satellites; ； in, is the GNSS comprehensive weight, is the GNSS signal quality indicator, is the number of satellites; Constructing multipath effect evaluation value To dynamically adjust the weight of the high-resistance UWB, the signal strength of the high-resistance UWB is , then the weight calculation formula of high robustness UWB is as follows: ； in, is the UWB comprehensive weight, is the UWB signal strength, is the multipath effect evaluation value; Through accelerometer noise and gyroscope noise Calculate the weight of INS , as shown in the following formula: ； in, is the comprehensive weight of INS, is the accelerometer noise, Gyroscope noise.

10. An intelligent unmanned vehicle positioning system in a coal mine, characterized in that: include: Inoue positioning module: Build a neural network to assist the GNSS and INS combined system to locate the unmanned vehicle; Downhole positioning module: Improve the robustness of the UWB positioning algorithm to obtain highly robust UWB and build a highly robust dual-layer fusion architecture of UWB and INS tightly combined system to locate the unmanned vehicle; Transition area positioning module: Build a weighted fusion high-resistance UWB, GNSS and INS combined system to locate the unmanned vehicle.

Images