UWB high-precision three-dimensional positioning method and device based on multi-information evolution matching
By adopting multi-information evolution matching and grid screening methods in UWB three-dimensional positioning technology, combined with LSTM model and fingerprint library, the problem of insufficient positioning accuracy and real-time in the existing technology is solved, and a more efficient UWB three-dimensional positioning effect is achieved.
Patent Information
- Application Number
- CN202311754092.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-20
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2043-12-20
AI Technical Summary
The existing UWB three-dimensional positioning technology has problems of high complexity and poor accuracy, and has failed to effectively use the multi-information evolution state for three-dimensional positioning grid screening, resulting in low positioning accuracy and poor real-time performance.
Using a method based on multi-information evolution matching, a three-dimensional positioning fingerprint library is constructed by dividing spatial grid areas, and a LSTM model with fusion attention mechanism is used to predict the evolution state of multi-information, combining grid confidence and fingerprint adaptation for grid screening and positioning results optimization.
It improves the accuracy and real-time performance of UWB three-dimensional positioning, reduces the complexity of the positioning process, enhances the anti-interference ability, and can better cope with multipath effects and signal interference in complex environments.
Smart Images

Figure CN117761620B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of UWB three-dimensional positioning, and in particular to a UWB high-precision three-dimensional positioning method and device based on multi-information evolution matching. Background Art
[0002] The rapid development of the information age has made location information an important part of social development and national economic production, and it has been widely used in fields such as communications, transportation, and power systems. With the ever-changing spatial environment, people's demand for positioning is increasing, and the requirements for positioning accuracy are also increasing. Ultra-Wide Band (UWB) three-dimensional positioning technology has been widely used in the field of multi-scenario positioning due to its extremely high resolution and anti-interference capabilities, and has played an outstanding role in production supervision, intelligent transportation, and augmented reality. However, due to the diversity and complexity of signal propagation paths, there is the influence of multi-path interference in the propagation process of UWB signals, and UWB three-dimensional positioning technology faces the challenges of difficult high-precision positioning and low positioning reliability. In order to ensure the reliability of UWB three-dimensional positioning, reduce positioning errors, and achieve high-precision three-dimensional positioning, it is necessary to effectively utilize the position evolution information contained in the UWB signal.
[0003] At present, there have been some studies on UWB 3D positioning technology:
[0004] The patent "A UWB high-precision positioning method and system" proposes a UWB high-precision positioning method and system. First, the one-way radio electromagnetic wave transmission time difference obtained by each base station in the current single ranging phase and the communication with the tag, the coordinate information of each base station, and the antenna delay compensation curve of each base station and the tag communication obtained in advance are obtained; then, the distance between each base station and the tag is calculated based on the obtained data; finally, the real-time position coordinates of the tag are determined based on the calculated distance between each base station and the tag.
[0005] The patent "A UWB array fusion high-precision positioning method and system" proposes a UWB array fusion high-precision positioning method and system. First, provide a UWB array active tag and deploy the base station at a fixed position with known coordinates; second, measure the time it takes for the signal of each base station to reach the UWB array active tag, and use the ranging equation to convert the time it takes for the signal of each base station to reach the UWB array active tag into distance; then, measure the distance from the three base stations to the UWB array active tag; finally, use the extended Kalman filter based on state distance constraints to calculate the location information of the UWB array active tag.
[0006] However, the existing UWB three-dimensional positioning technology still has the following two shortcomings: First, it does not consider the evolution state of multiple information for three-dimensional positioning grid screening, resulting in a high complexity of the three-dimensional positioning process, and the accuracy and real-time performance of three-dimensional positioning are poor, and it cannot meet the user's positioning needs. Second, it does not consider the integration of multiple algorithms for UWB high-precision three-dimensional positioning optimization, resulting in low accuracy of UWB three-dimensional positioning results. Therefore, it is urgent to design a UWB high-precision three-dimensional positioning technology and device based on multi-information evolution matching, fully consider UWB ranging information, signal strength, and time difference to reach the base station and other information, divide the space into multiple grid areas, build a three-dimensional positioning fingerprint library, match and screen the positioning grid based on the evolution state of multiple information, and combine the grid confidence and fingerprint adaptation to comprehensively determine the user's position, effectively improve the accuracy of the UWB three-dimensional positioning results, while reducing the complexity of the three-dimensional positioning process and improving the real-time performance of three-dimensional positioning. Summary of the invention
[0007] The purpose of the present invention is to provide a UWB high-precision three-dimensional positioning method based on multi-information evolution matching, which can not only screen the three-dimensional positioning grid to improve efficiency, but also integrate multiple algorithms to achieve user positioning, thereby enhancing the anti-interference ability of UWB three-dimensional positioning technology.
[0008] In order to achieve the above object, the present invention is implemented through the following technical solutions:
[0009] The UWB high-precision three-dimensional positioning method based on multi-information evolution matching includes the following steps:
[0010] Initialize each module and divide the space environment into multiple grid areas;
[0011] Build a high-precision three-dimensional positioning fingerprint library, with each grid area as a data collection area. Move the UWB tag within the grid area and collect basic data, including the ranging information, signal strength RSSI, and time difference TDOA of the UWB tag. Get the positioning result based on the basic data, and use the basic data and positioning result as the fingerprint information of each collection point. The fingerprint information of all collection points constitutes the three-dimensional positioning fingerprint library;
[0012] Multi-information evolution prediction: The base station obtains the detection data of the user's location at regular time slots. The detection data includes the ranging information of the user's positioning, the signal strength RSSI, and the time difference TDOA of the signal arriving at the base station. The LSTM multi-information evolution prediction model with the fusion attention mechanism is used to predict the evolution state of the multi-information and output the predicted multi-information evolution state.
[0013] 3D positioning grid screening, using the predicted multi-information evolution state matching to obtain the confidence of each grid area, and screening out the N grid areas with the largest grid confidence;
[0014] Based on the RSSI spherical positioning algorithm and the TDOA arrival time difference positioning algorithm, the corresponding positioning results are calculated respectively, and the positioning results are compared with the fingerprint information of N grids obtained by screening the three-dimensional positioning grid to obtain the fingerprint fitness. The comprehensive fitness of the screened N grid areas is calculated using the grid confidence and fingerprint fitness, and the grid area corresponding to the position fingerprint with the highest comprehensive fitness is selected as the user's final positioning result.
[0015] Furthermore, the method for obtaining the positioning result according to the basic data comprises the steps of:
[0016] At each collection point, select the collection data of K base stations that are not in the same plane and are closest to the collection point as basic data;
[0017] For the mth acquisition point in grid area n, based on UWB ranging information and signal strength RSSI, the spherical positioning technology is used to obtain the RSSI positioning result of the acquisition point;
[0018] Based on the time difference TDOA of the signal arriving at the base station, the acquisition point is positioned using the arrival time difference positioning technology to obtain the TDOA positioning result of the acquisition point.
[0019] Furthermore, the grid area n Fingerprint information It can be expressed as:
[0020]
[0021] in, is the UWB ranging fingerprint information of grid area n, is the UWB ranging fingerprint information of grid area n collected by the i-th base station, is the signal strength RSSI fingerprint information of grid area n, The signal strength RSSI fingerprint information collected by the i-th base station in grid area n, is the TDOA fingerprint information of the signal arrival time difference of the base station in grid area n, The TDOA fingerprint information of the signal arrival time difference of the grid area n collected by the i-th base station, is the fingerprint information of the positioning coordinates result based on RSSI in grid area n, is the three-dimensional positioning coordinate of grid area n based on RSSI, is the fingerprint information of the positioning coordinate result based on TDOA in grid area n, is the three-dimensional positioning coordinate of grid area n based on TDOA, i∈[1,K];
[0022] The fingerprints of the grid areas constitute the fingerprint library FL:
[0023]
[0024] Furthermore, the LSTM multi-information evolution prediction model integrating the attention mechanism is used to predict the evolution state of multi-information, including the following steps:
[0025] Assume that the length of the multi-information evolution time window is X, input the multi-information evolution state set of the [tX-1, t-1] time slot into the model and perform cascade;
[0026] The attention mechanism is introduced to learn the importance vector of each dimension in the evolution of multi-information through network training, and the importance vector is used to weight the input multi-information evolution state to obtain the attention result;
[0027] The de-cascaded attention results are fed into the LSTM multi-information evolution prediction model that integrates the attention mechanism;
[0028] The model outputs a set of multi-information evolution states of the predicted user in time slot t.
[0029] Furthermore, the multi-information evolution state includes the multi-information mean and the multi-information variance.
[0030] Furthermore, D(t) is defined as the ranging information of the user in time slot t, R(t) is the signal strength of the user in time slot t, and ΔT(t) is the time difference of the signal arriving at the base station in time slot t. Therefore, in the time period [tX-1, t-1], the evolution state set of ranging information, signal strength, and time difference of signal arriving at the base station can be expressed as:
[0031] {D(tX-1),D(tX),…,D(t-1)}
[0032] {R(tX-1),R(tX),…,R(t-1)}
[0033] {ΔT(tX-1),ΔT(tX),…,ΔT(t-1)}
[0034] The evolution state set of user multi-information from time slot tX-1 to time slot t-1 includes:
[0035] The mean evolution state set of ranging information {D avg (tX-1),D avg (tX),…,D avg (t-1)};
[0036] The variance evolution state set of ranging information {D var (tX-1),D var(tX),…,D var (t-1)};
[0037] The mean evolution state set of signal strength {R avg (tX-1),R avg (tX),…,R avg (t-1)};
[0038] The variance evolution state set of signal strength {R var (tX-1),R var (tX),…,R var (t-1)};
[0039] The mean evolution state set of time difference {ΔT avg (tX-1),ΔT avg (tX),…,ΔT avg (t-1)};
[0040] The variance evolution state set of time difference {ΔT var (tX-1),ΔT var (tX),…,ΔT var (t-1)}.
[0041] Among them, D avg (t-1) represents the mean value set of user t-1 time slot ranging information detected by each base station, D var (t-1) represents the variance set of the user t-1 time slot ranging information detected by each base station, R avg (t-1) represents the mean value set of the signal strength of user t-1 time slot detected by each base station, R var (t-1) represents the variance set of the signal strength of user t-1 time slot detected by each base station, ΔT avg (t-1) represents the mean value set of the arrival time difference of the user t-1 time slot signal detected by each base station, ΔT var (t-1) represents the variance set of the arrival time difference of the user t-1 time slot signal detected by each base station;
[0042] The LSTM multi-information evolution prediction model with fusion attention mechanism is used to predict the evolution state of multi-information and then output the predicted multi-information mean set and multi-information variance set of the user in time slot t:
[0043] The predicted mean value set of ranging information of each base station in time slot t The predicted variance set of ranging information of each base station in time slot t The predicted signal strength mean value set of each base station in time slot t The predicted signal strength variance of each base station in time slot t The predicted time difference mean value set of each base station in time slot t The predicted time difference variance set of each base station in time slot t
[0044] Furthermore, using the predicted multi-information evolution state matching to obtain the confidence of each grid area includes the following steps:
[0045] For the grid area n, its fingerprint is matched with the multi-information evolution state to solve its confidence, which can be expressed as:
[0046]
[0047] Among them, ξ1, ξ2, and ξ3 are the weights of the confidence obtained based on the ranging information, the weights of the confidence obtained based on the signal strength RSSI, and the weights of the confidence obtained based on the signal arrival time difference TDOA, respectively. represents the confidence that the user in time slot t is in grid area n based on ranging information matching, It represents the confidence that the user in time slot t is in grid area n based on signal strength RSSI matching. represents the confidence that the user in time slot t is in grid area n based on signal arrival time difference TDOA matching, K is the number of base stations, is the ranging information that may be detected by the kth base station in the predicted time slot t, is the corresponding confidence level; is the signal strength that may be detected by the kth base station in the predicted time slot t, is the corresponding confidence level; is the predicted signal arrival time difference that may be detected by the kth base station in time slot t, The corresponding confidence level.
[0048] Furthermore, the calculation method of fingerprint suitability includes the steps of:
[0049] The user's positioning result P is calculated based on the spherical positioning technology based on RSSI information and the arrival time difference positioning technology based on TDOA information. R (t) = (x R (t),y R (t),z R (t)) and P D (t) = (x D (t),y D (t),z D (t)), x R (t),y R (t),z R (t) is the three-dimensional positioning coordinate obtained by the spherical positioning technology based on RSSI information, xD (t),y D (t),z D (t) is the three-dimensional positioning coordinate obtained by the time difference of arrival positioning technology based on TDOA information;
[0050] The fingerprint adaptation degree of the user in time slot t and grid area n is:
[0051]
[0052] in, is the RSSI positioning fingerprint corresponding to the nth grid area in the fingerprint library, and θ1 is the adaptation weight of the RSSI positioning method; is the TDOA positioning fingerprint corresponding to the nth grid in the fingerprint library, and θ2 is the adaptation weight of the TDOA positioning method.
[0053] Furthermore, the calculation method of the comprehensive fitness includes the steps of:
[0054] The comprehensive adaptability ψ between user t time slot and grid area n n (t) can be expressed as:
[0055]
[0056] Among them, κ is the exponential weighting factor, C n (t) is the confidence level of grid area n at time slot t.
[0057] A UWB high-precision three-dimensional positioning device based on multi-information evolution matching, characterized in that it includes a data acquisition module, a three-dimensional positioning fingerprint library module, a multi-information evolution prediction module, a grid confidence calculation module, a three-dimensional positioning grid screening module, a spherical positioning module, a time difference positioning module, a fingerprint matching degree calculation module, a comprehensive matching degree calculation module and a high-precision three-dimensional positioning module;
[0058] There are multiple base stations in the space to achieve full UWB signal coverage, and at least four base stations are not in the same plane. The data acquisition module is used to achieve full UWB signal coverage and divide the space environment into grid areas. Each grid area includes a mobile UWB tag and a user. It is used to collect basic data in each grid area and transmit the basic data to the three-dimensional positioning fingerprint library module;
[0059] The three-dimensional positioning fingerprint library module is connected to the data acquisition module signal, and is used to construct and store fingerprint information of each grid area based on the basic data collected by the data acquisition module;
[0060] The multi-information evolution prediction module is used to receive the detection data of the base station on the user's position within the time window, and based on the detection data within the time window, use the LSTM multi-information evolution prediction model with fusion attention mechanism to predict the evolution state of the multi-information, and transmit the prediction results to the grid confidence calculation module;
[0061] The grid confidence calculation module is connected to the multi-information evolution prediction module by signal, and is used to calculate the confidence of each grid area based on the evolution prediction state of the multi-information evolution prediction module, and transmit the result to the three-dimensional positioning grid screening module;
[0062] The three-dimensional positioning grid screening module is connected to the grid confidence calculation module by signal, and is used to screen out N grid areas with the largest grid confidence based on the confidence of each grid area calculated by the grid confidence calculation module;
[0063] The spherical positioning module is used to obtain the positioning result based on the RSSI information based on the user signal strength RSSI detected in real time by using the spherical positioning technology, and send the user's RSSI positioning result to the fingerprint matching degree calculation module;
[0064] The time difference positioning module is used to obtain the positioning result based on the TDOA information based on the arrival time difference of the user signal detected in real time, and send the user's TDOA positioning result to the fingerprint matching degree calculation module;
[0065] The fingerprint matching degree calculation module is connected with the spherical positioning module and the time difference positioning module signal, and is used to calculate the fingerprint fitness corresponding to the N grids screened by the three-dimensional positioning grid screening module based on the positioning results of the spherical positioning module and the time difference positioning module, using the RSSI positioning results and TDOA positioning results of the user, and transmit the results to the comprehensive matching degree calculation module;
[0066] The comprehensive matching degree calculation module is connected to the fingerprint matching degree calculation module by signal, and is used to calculate the comprehensive fitness of the N grid areas selected by using the grid confidence and the fingerprint fitness based on the results of the fingerprint matching degree calculation module and the grid confidence calculation module, and transmit the results to the high-precision three-dimensional positioning module;
[0067] The high-precision three-dimensional positioning module is connected to the comprehensive matching degree calculation module to select the location fingerprint with the highest comprehensive adaptability according to the results of the comprehensive matching degree calculation module, and the corresponding grid area is the user's location.
[0068] The advantages of the present invention are:
[0069] (1) The present invention proposes a three-dimensional positioning grid screening method based on multi-information evolution matching. The space is divided into several grid areas according to the requirements of space size information and positioning accuracy. The UWB positioning result is calculated based on UWB ranging information, signal strength RSSI and time difference TDOA of arrival at the base station to form a fingerprint library of each grid area. At the same time, the LSTM multi-information evolution prediction model with integrated attention mechanism is used to predict the multi-information evolution state, predict the multi-information mean and multi-information variance of the user's current time slot, and calculate the grid confidence of each grid area based on the multi-information matching of ranging information, signal strength and signal arrival time difference evolution state. The positioning grids are matched and screened accordingly to screen out several grid areas with the largest confidence, which effectively solves the problem that traditional three-dimensional positioning technology is difficult to cope with complex and changeable positioning environments, reduces the complexity of the three-dimensional positioning process, and enhances the real-time performance of three-dimensional positioning.
[0070] (2) The present invention proposes a UWB high-precision three-dimensional positioning technology based on multi-algorithm fusion, and uses multiple positioning algorithms to perform high-precision positioning of the user's position. First, the user's signal strength and the time difference of arrival at the base station are measured, and the positioning results are calculated using the RSSI spherical positioning algorithm and the TDOA arrival time difference positioning algorithm respectively. Then, the fingerprint adaptation of the user in each grid in the three-dimensional space is determined by combining the positioning results of the multiple algorithms. Finally, the comprehensive adaptation of the user and each grid area is calculated by comprehensively considering the grid confidence and the fingerprint adaptation. The grid area with the largest comprehensive adaptation is the user's location, thereby achieving high-precision three-dimensional positioning, fully considering the correlation of the user positioning results of each time slot, and exploring the influence of the user's past time slot location on the user positioning of the current time slot, effectively improving the accuracy of UWB three-dimensional positioning, and enhancing the anti-interference ability of UWB three-dimensional positioning technology, so as to better deal with the multipath effect and signal interference in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS
[0071] Figure 1 This is a flow chart of the UWB high-precision three-dimensional positioning method based on multi-information evolution matching;
[0072] Figure 2 It is an LSTM multi-information evolution prediction model integrating attention mechanism;
[0073] Figure 3 It is a UWB high-precision three-dimensional positioning device based on multi-information evolution matching. DETAILED DESCRIPTION
[0074] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.
[0075] Example 1
[0076] This embodiment proposes a UWB high-precision three-dimensional positioning method based on multi-information evolution matching. For the specific process, please refer to Figure 1 Firstly, a three-dimensional positioning grid screening method based on multi-information evolution matching is proposed. The grid area is divided, and a fingerprint library of three-dimensional positioning is constructed. Multi-information evolution matching is realized to screen out multiple grid areas with the highest confidence. Then, according to the UWB high-precision three-dimensional positioning method with multi-algorithm fusion, the corresponding positioning results are calculated based on the RSSI spherical positioning algorithm and the TDOA arrival time difference positioning algorithm, and the user's three-dimensional positioning is realized by combining the grid confidence and fingerprint adaptation.
[0077] The specific implementation steps are as follows:
[0078] S1: Three-dimensional positioning grid screening method based on multi-information evolution matching
[0079] Firstly, the UWB ranging information, signal strength RSSI and time difference TDOA of arrival at the base station are used to form the location information of each spatial grid, and a three-dimensional positioning fingerprint library is constructed based on this. Secondly, the long short-term memory network (LSTM) model with integrated attention mechanism is used to predict the multi-information evolution state, and the multi-information mean and multi-information variance of the user's current time slot are predicted. Finally, the grid confidence of each spatial grid is calculated based on the multi-information evolution state, and several grid areas with the largest confidence are screened out.
[0080] S1.1: Construction of high-precision 3D positioning fingerprint library
[0081] According to the space size information and positioning accuracy requirements, the space is divided into The UWB tag is moved in the grid area and data is collected. The collection position and number of collection points are reasonably set according to the difference of each grid area. The collected data includes UWB ranging information d, signal strength RSSI and time difference TDOA of signal arrival at base station.
[0082] At each collection point, the collection data of K base stations that are not in the same plane and are closest to the collection point are selected as basic data. For the mth collection point in grid area n, the RSSI positioning result of the collection point is obtained using spherical positioning technology based on UWB ranging information and signal strength RSSI; based on the time difference of signal arrival at the base station TDOA, the collection point is located using the arrival time difference positioning technology to obtain the TDOA positioning result of the collection point.
[0083] The basic data and positioning results of each collection point are used as the fingerprint information of the collection point. The collection point with the smallest separation from the mean point in each grid area is selected, and the fingerprint information of the collection point is used as the fingerprint of the grid area where it is located. Fingerprint of grid area n It can be expressed as:
[0084]
[0085] in, For the grid area n UWB ranging fingerprint information, is the UWB ranging fingerprint information of grid area n collected by the i-th base station, is the signal strength RSSI fingerprint information of grid area n, The signal strength RSSI fingerprint information collected by the i-th base station in grid area n, is the TDOA fingerprint information of the signal arrival time difference of the base station in grid area n, The TDOA fingerprint information of the signal arrival time difference of the grid area n collected by the i-th base station, For the grid area n Fingerprint information of positioning coordinate results based on RSSI, For the grid area n Three-dimensional positioning coordinates based on RSSI, is the fingerprint information of the positioning coordinate result based on TDOA in grid area n, is the 3D positioning coordinate of grid area n based on TDOA, i∈[1,K].
[0086] The fingerprints of all grid areas constitute the fingerprint library FL, which can be expressed as:
[0087]
[0088] S1.2: Prediction of multi-information evolution status
[0089] The evolution of user location information is not a purely random process. The location information of the user at time t is often related to the location information at the previous time. The present invention uses LSTM with fusion attention mechanism to predict the evolution state of the user's location information. The predicted data includes the user's ranging information, signal strength RSSI and time difference TDOA of the signal arriving at the base station.
[0090] The base station obtains the user's positioning information every other time slot. Assuming the length of the multi-information evolution time window is X, the present invention predicts the multi-information evolution state of the user in the t time slot based on the user's ranging information, signal strength RSSI and time difference TDOA arriving at the base station in the [tX-1, t-1] time slot.
[0091] The base station detects the user's location status and can obtain the ranging information of the user's positioning, signal strength RSSI, and the time difference TDOA of the signal arriving at the base station. The ranging information of the user's positioning in the t time slot, signal strength RSSI, and the time difference TDOA of the signal arriving at the base station can be expressed as:
[0092]
[0093] Where D(t) represents the ranging information of the user detected by the base station in time slot t, d1(t), d2(t),…,d K (t) represent the ranging information of the user detected by the 1st, 2nd, ..., Kth base stations in time slot t, R(t) represents the signal strength of the user detected by the base station in time slot t, r1(t), r2(t), ..., r K (t) represents the signal strength of user t time slot detected by the 1st, 2nd, ..., Kth base stations, ΔT(t) represents the time difference of the signal of user t time slot detected by the base station arriving at the base station, Δt1(t), Δt2(t), ..., Δt K (t) represents the time difference of the signal arrival at the base station of user t time slot detected by the 1st, 2nd, ... Kth base stations. Therefore, in the time period [tX-1, t-1], the evolution state set of ranging information, signal strength, and signal arrival time difference can be expressed as {D(tX-1), D(tX), ..., D(t-1)}, {R(tX-1), R(tX), ..., R(t-1)} and {ΔT(tX-1), ΔT(tX), ..., ΔT(t-1)} respectively.
[0094] Please refer to Figure 2 The present invention uses the LSTM multi-information evolution prediction model integrating the attention mechanism to predict the evolution state of the multi-information of the user t time slot, wherein the evolution state set of the multi-information of the user from the tX-1 time slot to the t-1 time slot includes:
[0095] (1) The mean evolution state set of ranging information {D avg (tX-1),D avg (tX),…,D avg (t-1)};
[0096] (2) The variance evolution state set of ranging information {D var (tX-1),D var (tX),…,D var (t-1)};
[0097] (3) The mean evolution state set of signal strength {R avg (tX-1),R avg (tX),…,Ravg (t-1)};
[0098] (4) The variance evolution state set of signal strength {R var (tX-1),R var (tX),…,R var (t-1)};
[0099] (5) The mean evolution state set of time difference {ΔT avg (tX-1),ΔT avg (tX),…,ΔT avg (t-1)};
[0100] (6) The variance evolution state set of time difference {ΔT var (tX-1),ΔT var (tX),…,ΔT var (t-1)}.
[0101] in, represents the mean value set of user t-1 time slot ranging information detected by the 1st, 2nd, ..., Kth base stations, represents the variance set of user t-1 time slot ranging information detected by the 1st, 2nd, ..., Kth base stations, represents the mean value set of the signal strength of user t-1 time slot detected by the 1st, 2nd, ..., Kth base stations, represents the variance set of the signal strength of user t-1 time slot detected by the 1st, 2nd, ..., Kth base stations, represents the mean value set of the arrival time difference of the user t-1 time slot signal detected by the 1st, 2nd, ..., Kth base stations, Represents the variance set of the arrival time difference of user t-1 time slot signal detected by the 1st, 2nd, ...Kth base stations.
[0102] The LSTM multi-information evolution prediction process integrating the attention mechanism is described as follows:
[0103] First, the multi-information mean evolution state set and variance evolution state set of the time slot [tX-1, t-1] are input to the model and cascaded; secondly, the attention mechanism is introduced, and the importance of each dimension in the multi-information evolution is learned through network training. The sign vector is used to weight the input multi-information mean evolution state and multi-information variance evolution state to obtain the attention result; then, the de-cascaded attention result is sent to the LSTM neural network; finally, the model outputs the predicted multi-information mean set and multi-information variance set of the user in the t time slot, including:
[0104] (1) The predicted mean value set of ranging information of the 1st, 2nd, ... Kth base stations in time slot t
[0105] (2) The predicted variance set of ranging information of the 1st, 2nd, ... Kth base stations in time slot t
[0106] (3) The predicted signal strength mean set of the 1st, 2nd, ... Kth base stations in time slot t
[0107] (4) The predicted signal strength variance set of the 1st, 2nd, ... Kth base stations at time slot t
[0108] (5) The predicted time difference mean set of the 1st, 2nd, ... Kth base stations in time slot t
[0109] (6) The predicted time difference variance set of the 1st, 2nd, ... Kth base stations in time slot t
[0110] Therefore, the multi-information mean set of t time slots obtained by joint prediction is and multi-information variance set According to the distribution it satisfies, the predicted t-time slot multi-information evolution state is solved:
[0111]
[0112] in, is the ranging information that may be detected by the kth base station in time slot t, is the corresponding confidence level; is the signal strength that the kth base station may detect in time slot t, is the corresponding confidence level; is the signal arrival time difference that may be detected by the kth base station in time slot t, The corresponding confidence level.
[0113] S1.3: 3D positioning grid screening method based on multi-information evolution matching grid confidence
[0114] The present invention combines ranging information matching, signal strength RSSI matching and signal arrival time difference TDOA matching, and uses multi-information evolution matching to obtain the confidence C of each grid area. n (t), and based on this, select the N grid areas with the largest confidence. n , match its fingerprint with the multi-information evolution state to solve its confidence, which can be expressed as:
[0115]
[0116] Among them, ξ1, ξ2, and ξ3 are respectively the weight of the confidence obtained based on the ranging information, the weight of the confidence obtained based on the signal strength RSSI, and the weight of the confidence obtained based on the signal arrival time difference TDOA. represents the confidence that the user in time slot t is in grid area n based on ranging information matching, Indicates that the user in time slot t obtained based on signal strength RSSI matching is in the grid area n The confidence level, Indicates that the user in time slot t obtained based on signal arrival time difference TDOA matching is in the grid area n confidence level.
[0117] S2: UWB high-precision 3D positioning based on multi-algorithm fusion
[0118] First, according to the RSSI and TDOA information at time t, the corresponding positioning results are calculated based on the RSSI spherical positioning algorithm and the TDOA arrival time difference positioning algorithm respectively; secondly, the positioning results are compared with the fingerprint information of N grids obtained by the three-dimensional positioning grid screening method based on the multi-information evolution matching grid confidence to obtain the fingerprint adaptation degree; finally, the most suitable position fingerprint is selected based on the comprehensive grid confidence and fingerprint adaptation degree to realize the positioning of the user. The specific introduction is as follows:
[0119] S2.1: Fingerprint fitness calculation method based on multi-algorithm fusion
[0120] The user's positioning result P is calculated based on the spherical positioning technology based on RSSI information and the arrival time difference positioning technology based on TDOA information. R (t) = (x R (t),y R (t),z R (t)) and P D (t) = (x D (t),y D (t),z D (t)), and calculate the fingerprint fitness corresponding to the N grids screened by the three-dimensional positioning grid screening method based on the multi-information evolution matching grid confidence.
[0121] Therefore, the results of RSSI positioning method and TDOA positioning method are comprehensively considered and compared with the data in the fingerprint library to obtain the fingerprint adaptation degree A of the user in time slot t and grid area n. n (t), is calculated as follows:
[0122]
[0123] in, is the RSSI positioning fingerprint corresponding to the nth grid area in the fingerprint library, and θ1 is the adaptation weight of the RSSI positioning method; is the TDOA positioning fingerprint corresponding to the nth grid in the fingerprint library, and θ2 is the adaptation weight of the TDOA positioning method.
[0124] S2.2: Three-dimensional positioning method based on fusion of grid confidence and fingerprint adaptation
[0125] In order to effectively improve the accuracy of positioning, the present invention proposes a comprehensive adaptation three-dimensional positioning method based on multiple algorithms, based on grid confidence and fingerprint adaptation, to select the most suitable position fingerprint based on the comprehensive grid confidence and fingerprint adaptation to achieve positioning of the user. n (t) can be expressed as:
[0126]
[0127] Among them, κ is an exponential weighting factor. The comprehensive fitness of N grid areas is calculated in sequence, and the grid with the largest comprehensive fitness is the grid area corresponding to the user's location.
[0128] Example 2
[0129] This embodiment proposes a Figure 3 The UWB high-precision three-dimensional positioning device based on multi-information evolution matching shown in the figure includes a data acquisition module, a three-dimensional positioning fingerprint library module, a multi-information evolution prediction module, a grid confidence calculation module, a three-dimensional positioning grid screening module, a spherical positioning module, a time difference positioning module, a fingerprint matching degree calculation module, a comprehensive matching degree calculation module and a high-precision three-dimensional positioning module.
[0130] Data acquisition module: There are multiple base stations in the space to achieve full UWB signal coverage, and at least four base stations are not in the same plane. The data acquisition module divides the space environment into grids. UWB tags and users move in each grid, and multiple sets of data are collected in each area. Each set of data includes ranging information, signal strength RSSI and time difference TDOA to the base station, and the data is transmitted to the three-dimensional positioning fingerprint library module.
[0131] Three-dimensional positioning fingerprint library module: Based on the information collected by the data acquisition module, it builds and stores the fingerprint information of each grid area.
[0132] Multi-information evolution prediction module: Based on the detection data within the time window, the LSTM multi-information evolution prediction model with integrated attention mechanism is used to predict the evolution state of multi-information, and the prediction results are transmitted to the grid confidence calculation module.
[0133] Grid confidence calculation module: Based on the evolution prediction state of the multi-information evolution prediction module, the confidence of each grid area is calculated and the result is transmitted to the three-dimensional positioning grid screening module.
[0134] 3D positioning grid screening module: based on the confidence of each grid area calculated by the grid confidence calculation module, screen out N grid areas with the largest grid confidence.
[0135] Spherical positioning module: Based on the real-time detected user signal strength RSSI, the spherical positioning technology is used to obtain the positioning result based on RSSI information, and the user's RSSI positioning result is sent to the fingerprint matching degree calculation module.
[0136] Time difference positioning module: Based on the real-time detected user signal arrival time difference, the arrival time difference positioning technology is used to obtain the positioning result based on TDOA information, and the user's TDOA positioning result is sent to the fingerprint matching degree calculation module.
[0137] Fingerprint matching degree calculation module: Based on the positioning results of the spherical positioning module and the time difference positioning module, the user's RSSI positioning results and TDOA positioning results are used to calculate the fingerprint adaptation degree corresponding to the N grids screened out by the three-dimensional positioning grid screening module, and the results are transmitted to the comprehensive matching degree calculation module.
[0138] Comprehensive matching degree calculation module: Based on the results of the fingerprint matching degree calculation module and the grid confidence degree calculation module, the comprehensive fitness of the N selected grid areas is calculated using the grid confidence degree and fingerprint fitness degree, and the results are transmitted to the high-precision three-dimensional positioning module.
[0139] High-precision three-dimensional positioning module: According to the results of the comprehensive matching degree calculation module, the location fingerprint with the highest comprehensive adaptability is selected, and its corresponding grid area is the user's location.
[0140] The above device is used to perform UWB high-precision three-dimensional positioning based on multi-information evolution matching. The specific steps are as follows.
[0141] S1: Initialize each module and divide the space environment into grids.
[0142] S2: The data acquisition module collects multiple sets of data in each area according to the requirements of spatial size information and positioning accuracy. Each set of data includes ranging information, signal strength RSSI and time difference TDOA when arriving at the base station, and transmits the data to the three-dimensional positioning fingerprint library module to build a three-dimensional positioning fingerprint library module. The three-dimensional positioning fingerprint library module stores the fingerprint information of each grid area.
[0143] S3: The multi-information evolution prediction module receives the detection data of the base station on the user's location within the time window, uses the LSTM multi-information evolution prediction model with integrated attention mechanism to predict the evolution state of multi-information, and transmits the multi-information mean and multi-information variance to the grid confidence calculation module.
[0144] S4: The grid confidence calculation module calculates the confidence C of each grid area based on the evolution prediction state of the multi-information evolution prediction module. n (t), the result is transmitted to the three-dimensional positioning grid screening module, and the three-dimensional positioning grid screening module screens out N grid areas with the largest grid confidence based on the confidence of each grid area calculated by the grid confidence calculation module.
[0145] S5: The spherical positioning module calculates the positioning result P based on the RSSI information based on the real-time detected user signal strength RSSI R (t), and the user's RSSI positioning result is sent to the fingerprint matching degree calculation module. The time difference positioning module calculates the positioning result P based on the TDOA information based on the real-time detected user signal arrival time difference. D (t), and send the user’s TDOA positioning result to the fingerprint matching degree calculation module.
[0146] S6: The fingerprint matching degree calculation module calculates the fingerprint matching degree A corresponding to the N grids selected by the three-dimensional positioning grid selection module based on the RSSI positioning result and the TDOA positioning result of the user. n (t).
[0147] S7: Based on the results of the fingerprint matching calculation module and the grid confidence calculation module, the comprehensive matching degree calculation module uses the grid confidence and fingerprint matching degree to calculate the comprehensive matching degree ψ of the selected N grid areas n (t), and transmit the result to the high-precision three-dimensional positioning module. The high-precision three-dimensional positioning module selects the location fingerprint with the highest comprehensive fitness according to the result of the comprehensive matching degree calculation module, and uses the grid area corresponding to the location fingerprint as the user's final positioning result.
[0148] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A UWB high-precision three-dimensional positioning method based on multi-information evolution matching, characterized in that: Includes steps: Initialize each module and divide the space environment into multiple grid areas; Build a high-precision three-dimensional positioning fingerprint library, with each grid area as a data collection area. Move the UWB tag within the grid area and collect basic data, including the ranging information, signal strength RSSI, and time difference TDOA of the UWB tag. Get the positioning result based on the basic data, and use the basic data and positioning result as the fingerprint information of each collection point. The fingerprint information of all collection points constitutes the three-dimensional positioning fingerprint library; The method for obtaining positioning results according to basic data comprises the steps of: At each collection point, select the collection data of K base stations that are not in the same plane and are closest to the collection point as basic data; For the mth collection point in grid area n, the RSSI positioning result of the collection point is obtained using spherical positioning technology based on the ranging information and signal strength RSSI of the UWB tag; Based on the time difference TDOA of the signal arriving at the base station, the acquisition point is located using the arrival time difference positioning technology to obtain the TDOA positioning result of the acquisition point; Multi-information evolution prediction: The base station obtains the detection data of the user's location at regular time slots. The detection data includes the ranging information of the user's positioning, the signal strength RSSI, and the time difference TDOA of the signal arriving at the base station. The LSTM multi-information evolution prediction model with the fusion attention mechanism is used to predict the evolution state of the multi-information and output the predicted multi-information evolution state. 3D positioning grid screening, using the predicted multi-information evolution state matching to obtain the confidence of each grid area, and screening out the N grid areas with the largest grid confidence; Based on the RSSI spherical positioning algorithm and the TDOA arrival time difference positioning algorithm, the corresponding positioning results are calculated respectively, and the positioning results are compared with the fingerprint information of the N grids obtained by screening the three-dimensional positioning grid to obtain the fingerprint fitness. The comprehensive fitness of the N grid areas screened out is calculated using the grid confidence and fingerprint fitness, and the grid area corresponding to the position fingerprint with the highest comprehensive fitness is selected as the user's final positioning result; The method for calculating the fingerprint suitability comprises the steps of: The user's positioning result is calculated based on the spherical positioning technology based on RSSI information and the arrival time difference positioning technology based on TDOA information. and P D (t) = (x D (t),y D (t),z D (t)), x R (t),y R (t),z R (t) is the three-dimensional positioning coordinate obtained by spherical positioning technology based on RSSI information, x D (t),y D (t),z D (t) is the three-dimensional positioning coordinate obtained by the time difference of arrival positioning technology based on TDOA information; The fingerprint adaptation degree of the user in time slot t and grid area n is: in, is the RSSI positioning fingerprint corresponding to the nth grid in the fingerprint library, and θ1 is the adaptation weight of the RSSI positioning method; is the TDOA positioning fingerprint corresponding to the nth grid in the fingerprint library, θ2 is the adaptation weight of the TDOA positioning method; The method for calculating the comprehensive fitness degree comprises the steps of: The comprehensive adaptability ψ between user t time slot and grid area n n (t) can be expressed as: Among them, κ is the exponential weighting factor, C n (t) is the grid area of time slot t n The confidence level is calculated and the comprehensive fitness of each grid area is calculated in turn. The grid with the largest comprehensive fitness is the grid area corresponding to the user's location.
2. The UWB high-precision three-dimensional positioning method based on multi-information evolution matching according to claim 1 is characterized in that: Fingerprint information of grid area n It can be expressed as: in, is the UWB ranging fingerprint information of grid area n, is the UWB ranging fingerprint information of grid area n collected by the i-th base station, is the signal strength RSSI fingerprint information of grid area n, The signal strength RSSI fingerprint information collected by the i-th base station in grid area n, is the TDOA fingerprint information of the signal arrival time difference of the base station in grid area n, The TDOA fingerprint information of the signal arrival time difference of the grid area n collected by the i-th base station, is the fingerprint information of the positioning coordinates result based on RSSI in grid area n, is the three-dimensional positioning coordinate of grid area n based on RSSI, is the fingerprint information of the positioning coordinate result based on TDOA in grid area n, is the three-dimensional positioning coordinate of grid area n based on TDOA, i∈[1,K]; The fingerprints of the grid areas constitute the fingerprint library FL:
3. The UWB high-precision three-dimensional positioning method based on multi-information evolution matching according to claim 1 is characterized in that: The LSTM multi-information evolution prediction model using the fusion attention mechanism predicts the evolution state of multi-information, including the following steps: Assume that the length of the multi-information evolution time window is X, input the multi-information evolution state set of the [tX-1, t-1] time slot into the model and perform cascade; The attention mechanism is introduced to learn the importance vector of each dimension in the evolution of multi-information through network training, and the importance vector is used to weight the input multi-information evolution state to obtain the attention result; The de-cascaded attention results are fed into the LSTM multi-information evolution prediction model that integrates the attention mechanism; The model outputs a set of multi-information evolution states of the predicted user in time slot t.
4. The UWB high-precision three-dimensional positioning method based on multi-information evolution matching according to claim 3 is characterized in that: The multi-information evolution state includes a multi-information mean and a multi-information variance.
5. The UWB high-precision three-dimensional positioning method based on multi-information evolution matching according to claim 4 is characterized in that: Define D(t) as the ranging information of the user in time slot t, R(t) as the signal strength of the user in time slot t, and ΔT(t) as the time difference of the signal arriving at the base station in time slot t. Therefore, in the time period [tX-1, t-1], the evolution state set of ranging information, signal strength, and time difference of signal arriving at the base station can be expressed as: {D(tX-1),D(tX),…,D(t-1)} {R(tX-1),R(tX),…,R(t-1)} {ΔT(tX-1),ΔT(tX),…,ΔT(t-1)} The evolution state set of user multi-information from time slot tX-1 to time slot t-1 includes: The mean evolution state set of ranging information {D avg (tX-1),D avg (tX),…,D avg (t-1)}; The variance evolution state set of ranging information {D var (tX-1),D var (tX),…,D var (t-1)}; The mean evolution state set of signal strength {R avg (tX-1),R avg (tX),…,R avg (t-1)}; The variance evolution state set of signal strength {R var (tX-1),R var (tX),…,R var (t-1)}; The mean evolution state set of time difference {ΔT avg (tX-1),ΔT avg (tX),…,ΔT avg (t-1)}; The variance evolution state set of time difference {ΔT var (tX-1),ΔT var (tX),…,ΔT var (t-1)}; Among them, D avg (t-1) represents the mean value set of user t-1 time slot ranging information detected by each base station, D var (t-1) represents the variance set of the user t-1 time slot ranging information detected by each base station, R avg (t-1) represents the mean value set of the signal strength of user t-1 time slot detected by each base station, R var (t-1) represents the variance set of the signal strength of user t-1 time slot detected by each base station, ΔT avg (t-1) represents the mean value set of the arrival time difference of the user t-1 time slot signal detected by each base station, ΔT var (t-1) represents the variance set of the arrival time difference of the user t-1 time slot signal detected by each base station; The LSTM multi-information evolution prediction model with fusion attention mechanism is used to predict the evolution state of multi-information and then output the predicted multi-information mean set and multi-information variance set of the user in time slot t: The predicted mean value set of ranging information of each base station in time slot t The predicted variance set of ranging information of each base station in time slot t The predicted signal strength mean value set of each base station in time slot t The predicted signal strength variance of each base station in time slot t The predicted time difference mean value set of each base station in time slot t The predicted time difference variance set of each base station in time slot t 6. The UWB high-precision three-dimensional positioning method based on multi-information evolution matching according to claim 1 is characterized in that: The confidence level of each grid area is obtained by matching the predicted multi-information evolution state, including the following steps: For the grid area n, its fingerprint is matched with the multi-information evolution state to solve its confidence, which can be expressed as: Among them, ξ1, ξ2, and ξ3 are the weights of the confidence obtained based on the ranging information, the weights of the confidence obtained based on the signal strength RSSI, and the weights of the confidence obtained based on the signal arrival time difference TDOA, respectively. represents the confidence that the user in time slot t is in grid area n based on ranging information matching, Indicates that the user in time slot t obtained based on signal strength RSSI matching is in the grid area n The confidence level, represents the confidence that the user in time slot t is in grid area n based on signal arrival time difference TDOA matching, K is the number of base stations, is the ranging information that may be detected by the k-th base station in the predicted time slot t, is the corresponding confidence level; is the signal strength that may be detected by the kth base station in the predicted time slot t, is the corresponding confidence level; is the predicted signal arrival time difference that may be detected by the kth base station in time slot t, The corresponding confidence level.
7. UWB high-precision three-dimensional positioning device based on multi-information evolution matching, characterized in that: It includes data acquisition module, 3D positioning fingerprint library module, multi-information evolution prediction module, grid confidence calculation module, 3D positioning grid screening module, spherical positioning module, time difference positioning module, fingerprint matching degree calculation module, comprehensive matching degree calculation module and high-precision 3D positioning module; The data acquisition module is used to achieve full coverage of UWB signals and divide the spatial environment into grid areas. Each grid area includes mobile UWB tags and users. It is used to collect basic data in each grid area and transmit the basic data to the three-dimensional positioning fingerprint library module. The basic data includes the ranging information of the UWB tag, the signal strength RSSI and the time difference TDOA when arriving at the base station. The positioning result is obtained based on the basic data. The basic data and the positioning result are used as the fingerprint information of each collection point. The fingerprint information of all collection points constitutes the three-dimensional positioning fingerprint library; The three-dimensional positioning fingerprint library module is connected to the data acquisition module signal, and is used to construct and store fingerprint information of each grid area based on the basic data collected by the data acquisition module; The multi-information evolution prediction module is used to receive the detection data of the base station on the user's position within the time window, and based on the detection data within the time window, use the LSTM multi-information evolution prediction model with fusion attention mechanism to predict the evolution state of the multi-information, and transmit the prediction results to the grid confidence calculation module; The grid confidence calculation module is connected to the multi-information evolution prediction module by signal, and is used to calculate the confidence of each grid area based on the evolution prediction state of the multi-information evolution prediction module, and transmit the result to the three-dimensional positioning grid screening module; The three-dimensional positioning grid screening module is connected to the grid confidence calculation module by signal, and is used to screen out N grid areas with the largest grid confidence based on the confidence of each grid area calculated by the grid confidence calculation module; The spherical positioning module is used to obtain the positioning result based on the RSSI information based on the user signal strength RSSI detected in real time by using the spherical positioning technology, and send the user's RSSI positioning result to the fingerprint matching degree calculation module; The time difference positioning module is used to obtain the positioning result based on the TDOA information based on the arrival time difference of the user signal detected in real time, and send the user's TDOA positioning result to the fingerprint matching degree calculation module; The fingerprint matching degree calculation module is connected with the spherical positioning module and the time difference positioning module signal, and is used to calculate the fingerprint fitness corresponding to the N grids screened by the three-dimensional positioning grid screening module based on the positioning results of the spherical positioning module and the time difference positioning module, using the RSSI positioning results and TDOA positioning results of the user, and transmit the results to the comprehensive matching degree calculation module; The comprehensive matching degree calculation module is connected to the fingerprint matching degree calculation module by signal, and is used to calculate the comprehensive fitness of the N grid areas selected by using the grid confidence and the fingerprint fitness based on the results of the fingerprint matching degree calculation module and the grid confidence calculation module, and transmit the results to the high-precision three-dimensional positioning module; The high-precision three-dimensional positioning module is connected to the comprehensive matching degree calculation module to select the location fingerprint with the highest comprehensive adaptability according to the results of the comprehensive matching degree calculation module, and the corresponding grid area is the user's location.
Citation Information
Patent Citations
Multidimensional positioning method and apparatus thereof
CN102360076A
Indoor positioning method based on logarithmic normal model
CN110557716A
WIFI indoor positioning method and system for large-scale complex scene
CN111726765A
Three-dimensional indoor positioning method and system based on ultra-wideband positioning and fingerprint positioning
CN116148766A
3D position tracking system using UWB
KR102009791B1
