Internet of things terminal multi-source adaptive fusion positioning system and method

Abstract

The application provides a kind of Internet of Things terminal multi-source adaptive fusion positioning system and method, belong to multi-source sensor information fusion positioning field, comprising: the original data of GNSS, IMU, barometer and magnetometer are synchronously collected;After the smooth processing of IMU data, multi-dimensional time domain features are extracted, and the state of pedestrian or vehicle is determined in real time according to the feature score;Under the same Kalman filtering framework, the observation vector is dynamically adjusted according to the state result and sensor availability, and the information of GNSS, INS / PDR, barometer and magnetometer is uniformly fused;According to the different inertial navigation calculation methods in pedestrian mode or vehicle mode, the inertial navigation system state model is switched, and the corresponding motion constraint is introduced, and the continuous positioning result is directly output through one-step Kalman time update and observation update.The application provides accurate positioning service and intelligent navigation experience for users.

Description

Technical Field

[0001] This invention belongs to the field of multi-source sensor information fusion positioning, and specifically relates to a multi-source adaptive fusion positioning system and method for Internet of Things terminals. Background Technology

[0002] With the increasing prevalence of the Internet of Things (IoT) and wearable devices, the public has placed higher demands on the positioning accuracy and stability of low-cost terminals. These devices operate in a variety of complex scenarios, including walking, cycling, and in-vehicle use. A single Global Navigation Satellite System (GNSS) or other navigation methods is insufficient to meet the positioning and navigation needs in these complex environments. Therefore, multi-source sensor fusion positioning has emerged as a key means to achieve stable navigation and positioning in multiple scenarios. GNSS chips, accelerometers, and gyroscopes are common sensor configurations in IoT / wearable devices, among which the combined navigation algorithm of GNSS and Inertial Navigation System (INS) is widely used. However, in complex environments such as urban canyons, GNSS signals are easily lost, and low-cost inertial measurement units (IMUs) accumulate errors rapidly due to sensor accuracy limitations. This necessitates the introduction of other sensors or constraint methods to provide more stable positioning services.

[0003] The essence of multi-source sensor fusion positioning lies in leveraging the complementary advantages of multiple sensors to improve positioning accuracy and robustness. For example, fusing sensors such as GNSS, INS, barometers, and magnetometers enables continuous positioning in indoor, outdoor, and environments with varying signal strengths, effectively compensating for the shortcomings of individual sensors and enhancing positioning accuracy and stability. However, multiple combined positioning algorithms consume significant computing and storage resources, which is difficult for IoT / wearable devices, which already have limited computing and storage resources, to handle. Therefore, when designing multi-source fusion positioning systems for IoT / wearable devices, it is crucial to consider the power consumption limitations of hardware processors and reduce computational load from the algorithm level to create a lightweight multi-source fusion positioning system. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a multi-source adaptive fusion positioning system and method for IoT terminals. The system comprises a GNSS module, an INS navigation module, a barometer and magnetometer module, a user status recognition module, and a multi-source information adaptive processing module. The GNSS, INS, and barometer / magnetometer modules acquire output data from sensor modules in IoT devices (such as mobile phones and watches), calculate corresponding pose information, and transmit the results to the multi-source information adaptive processing module. The user status recognition module identifies the user's status based on IMU data motion feature analysis, then uses Kalman filtering to achieve multi-source information fusion processing, constructs a unified observation model, and determines an adaptive adjustment coefficient matrix for the observations based on the user status and multi-sensor availability. This enables adaptive switching when fusing information from different sensors, while avoiding redundant calculations caused by switching between different positioning algorithms, thus improving computational efficiency. This multi-source adaptive fusion positioning system can switch between different positioning models for IoT terminal devices in pedestrian and vehicular scenarios, and can implement various combination algorithms such as GNSS / INS combination, GNSS / PDR combination, and GNSS / PDR / magnetometer / barometer combination. This invention can apply different positioning algorithms according to the motion state and positioning requirements in different scenarios, providing users with accurate positioning services and intelligent navigation experience.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] An IoT terminal multi-source adaptive fusion positioning system includes: a GNSS module for acquiring satellite navigation positioning information; an inertial navigation module for acquiring raw IMU data and providing INS mechanical orchestration and PDR processes; a user state recognition module for extracting multi-dimensional motion features based on IMU data and outputting pedestrian or vehicle status; a barometer and magnetometer module for providing elevation and heading assistance information; and a multi-source information adaptive processing module for receiving data from the above modules, dynamically adjusting the observation vector within the same observation model based on user status and sensor availability, and outputting the fusion positioning result through a single Kalman filter, thereby achieving seamless continuous operation of the positioning algorithm in different scenarios.

[0007] The present invention also provides a multi-source adaptive fusion positioning method for IoT terminals, comprising the following steps:

[0008] Step 1: Synchronously acquire raw data from GNSS, IMU, barometer, and magnetometer;

[0009] Step 2: After smoothing the IMU data, extract multi-dimensional temporal features and determine the status of pedestrians or vehicles in real time based on the feature scores.

[0010] Step 3: Under the same Kalman filter framework, dynamically adjust the observation vector based on the state results of Step 2 and sensor availability, and integrate GNSS, INS / PDR, barometer and magnetometer information.

[0011] Step 4: Switch the inertial navigation system state model according to the different inertial navigation calculation methods in pedestrian mode or vehicle mode, and introduce the corresponding motion constraints. Output continuous positioning results directly through one-step Kalman time update and observation update.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described multi-source adaptive fusion positioning method for an Internet of Things terminal.

[0013] The present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described multi-source adaptive fusion positioning method for an Internet of Things terminal.

[0014] Beneficial effects:

[0015] 1. In response to the user needs of IoT device application platforms, this invention integrates data from multiple sensors such as GNSS, IMU, barometer, and magnetometer. By comprehensively analyzing these data, pose information is calculated, which significantly improves the accuracy and stability of positioning.

[0016] 2. This invention uses an IMU-based multidimensional feature extraction and recognition method to distinguish the motion patterns of users in vehicles and pedestrians. It has advantages such as low computational load, high real-time performance, and no need for training data, and is easy to implement in IoT devices with limited resources.

[0017] 3. This invention employs the Kalman filtering algorithm for multi-source information fusion, fully considering the computational limitations of IoT devices. To this end, a unified observation model and adaptive adjustment coefficient matrix are constructed, enabling dynamic adjustment of observations based on user status and sensor availability, achieving adaptive switching between different sensor information fusion methods. This design effectively avoids redundant computations caused by switching between different positioning algorithms, significantly improving computational efficiency.

[0018] 4. This invention supports multiple combined positioning algorithms, including GNSS / INS combination, GNSS / PDR combination, and GNSS / PDR / magnetometer / barometer combination. Depending on different scenarios and motion states, the system can flexibly select the appropriate positioning algorithm to provide users with more accurate and intelligent positioning services. Attached Figure Description

[0019] Figure 1This is a schematic diagram of a multi-source adaptive fusion positioning system for an Internet of Things (IoT) terminal according to the present invention.

[0020] Figure 2 This is a flowchart of a multi-source adaptive fusion positioning method for IoT terminals according to the present invention.

[0021] Figure 3 This is a schematic flowchart of the GNSS / PDR / magnetometer / barometer combination method according to an embodiment of the present invention.

[0022] Figure 4 This is a schematic diagram of the GNSS / INS combination method according to an embodiment of the present invention.

[0023] Figure 5 This is a comparison chart of the positioning results using the present invention and the traditional vehicle-mounted combined positioning mode. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0025] This invention provides a multi-source adaptive fusion positioning system and method for IoT terminals, addressing user needs in IoT device application platforms. By combining GNSS, IMU, barometer, and magnetometer modules, it utilizes data from multiple sensors to calculate pose information, thereby improving the accuracy and stability of positioning.

[0026] like Figure 1 As shown, the IoT terminal multi-source adaptive fusion positioning system of the present invention includes:

[0027] The GNSS module is used to acquire raw GNSS observation information from IoT terminals and perform satellite navigation and positioning, i.e., GNSS positioning, outputting position, velocity, and corresponding observation quality information.

[0028] The Inertial Navigation (INS) module is used to acquire raw data from the IoT terminal's IMU and can perform INS mechanics orchestration and Pedestrian Dead Reckoning (PDR) processes. INS mechanics orchestration outputs the position, velocity, and attitude results calculated by INS, while PDR outputs the position and heading results.

[0029] The barometer and magnetometer modules are used to acquire barometer and magnetometer data from IoT terminals. The barometer can be used to obtain relative elevation information, and the magnetometer can be used to obtain relative heading information, thereby assisting in multi-source fusion positioning for motion state constraints.

[0030] The user status recognition module is used to identify the user's motion status. Through IMU data smoothing, feature extraction, and multi-dimensional recognition and judgment processes, it can identify the motion status of users, pedestrians, and vehicles, and transmit the recognition results to the INS navigation module and the multi-source information adaptive processing module.

[0031] The multi-source information adaptive processing module is used to perform multi-sensor information fusion. It achieves adaptive switching when fusing information from different sensors through a unified multi-source data comprehensive observation model and block-based state coefficients (i.e., the observation vector adaptive adjustment coefficient matrix), and obtains multi-source information adaptive positioning results.

[0032] Preferably, in the user state recognition module, user behavior recognition is achieved based on the differences in IMU motion characteristics between vehicle-mounted and pedestrian-mounted motions. Based on the differences in time-domain characteristics of IMU accelerometer and gyroscope data under pedestrian and vehicle-mounted states, multiple feature values ​​such as accelerometer peak ratio, variance, autocorrelation degree, and gyroscope commutation frequency are extracted as recognition criteria to construct a user state recognition model based on multi-dimensional IMU features.

[0033] Specifically, in the user state recognition module, a moving average filter is used to smooth the noise in the IMU accelerometer and gyroscope data to reduce the impact of sensor noise. The peak-to-peak ratio and variance features are then extracted from the smoothed IMU accelerometer data, as follows:

[0034] (1)

[0035] (2)

[0036] in, The peak ratio characteristic value of the accelerometer. Indicates the IMU epoch. The total number of IMU epochs for the feature extraction window. Indicates the first Z-axis acceleration of each epoch and These represent the maximum and average values ​​of the Z-axis acceleration, respectively. For the accelerometer variance eigenvalues, and These represent accelerations along the X and Y axes, respectively. and These are the average values ​​of the X-axis and Y-axis accelerations within the window epoch, respectively.

[0037] Specifically, the autocorrelation of the accelerometer data is calculated to identify periodic characteristics during walking. The autocorrelation index is obtained as follows:

[0038] (3)

[0039] (4)

[0040] (5)

[0041] in, This represents the hysteresis search time within the desired step frequency range, determined by combining the IMU sampling rate and pedestrian movement time. The value of , For accelerometer autocorrelation parameters, For delay The accelerometer autocorrelation function value at time 1. This is the accelerometer's self-power value. The resultant acceleration of the three-axis accelerometer. and This also represents the total number of IMU epochs and the feature extraction window IMU epochs.

[0042] Specifically, the motion characteristics of the gyroscope are used to assist in the determination of the user's motion state. The angular velocity commutation frequency and the angular velocity fluctuation variance are extracted as feature values, defined as follows:

[0043] (6)

[0044] (7)

[0045] in, This is the characteristic value of the gyroscope's commutation frequency. The window duration. This indicates that the value is 1 when the direction of the gyroscope's Z-axis angular velocity changes (i.e., changes from positive to negative). Indicates the first The angular velocity value of each epoch in the Z-axis direction; These are the eigenvalues ​​of the gyroscope's variance. This represents the average Z-axis angular velocity within the window. and This also represents the total number of IMU epochs and the feature extraction window IMU epochs.

[0046] Preferably, in the user state recognition module, a state recognition model based on multi-dimensional IMU features is constructed by integrating the aforementioned IMU motion time-domain features, as follows:

[0047] (8)

[0048] This involves dividing the IMU data into windows. For the first The state recognition score of each window For feature value index, The total number of eigenvalues ​​used. For the first The static weights of each feature represent the fundamental importance of that feature in the overall discrimination. The normalization function output for the j-th feature normalizes different feature values ​​to the interval [0,1]. The feature normalization function takes the following form:

[0049] (9)

[0050] Among them, for the feature normalization function , This is the original output of the j-th feature. and These are the minimum and maximum values ​​of the j-th feature in the empirical data, respectively. By standardizing the original features to the [0,1] interval, they are used for unified scoring across features.

[0051] Preferably, in the multi-source information adaptive processing module, different combination algorithms are switched according to the user status and the availability of multiple sensors. Specifically, in walking mode, GNSS and PDR positioning are used as the main positioning methods, and barometer and magnetometer information are used as auxiliary methods to constrain motion state; in cycling and vehicle modes, GNSS and INS navigation calculation are used as the main positioning methods, and zero speed constraint and heading constraint information are used as auxiliary methods to constrain motion state.

[0052] Preferably, in the multi-source information adaptive processing module, Kalman filtering is used to achieve the fusion of multi-source data, and a state-space model of multi-source fusion extended Kalman filtering and a multi-source data comprehensive observation model are constructed. The linearized model is as follows:

[0053] (10)

[0054] (11)

[0055] Where the subscripts k and k-1 represent the epoch numbers, and the system state parameters and system noise at time k-1 are respectively... and At time k, the system state parameters are Observations are The observation state matrix is The observed noise is ; This represents the system state transition matrix from epoch k-1 to epoch k.

[0056] Using the inertial navigation system as the main filter, the state parameter vector related to the inertial navigation system Includes three-dimensional position, velocity, attitude, and zero bias of accelerometers and gyroscopes:

[0057] (12)

[0058] In the formula, , and These are the position, velocity, and attitude error vectors, respectively. and These are the accelerometer zero bias and gyroscope zero bias error vectors, respectively.

[0059] Preferably, in the multi-source information adaptive processing module, pedestrian mode and vehicle mode adopt different inertial navigation calculation methods, namely PDR calculation method and INS mechanical arrangement calculation method, respectively, and the inertial navigation system state model switches according to the different inertial navigation calculation methods:

[0060] (13)

[0061] In the formula, This refers to the system state transition matrix corresponding to the position, velocity, and attitude parameters, which is switched according to different inertial navigation calculation methods. This is the state transition matrix corresponding to the zero bias error state parameters of the accelerometer and gyroscope. Given the state transition matrix between the position, velocity, and attitude state parameters and the zero bias error parameter, the system state model for position, velocity, and attitude, calculated using the PDR method, can be described as follows:

[0062] (14)

[0063] In the formula, , and In order to be in The derivatives of position error, velocity error, and attitude error in the navigation coordinate system. For the projection of the force in the n-system, For the attitude angle error based on the PSI angle model, Let be the rotation matrix from frame b (the carrier coordinate system) to frame n. This is the zero bias error vector of the accelerometer. and These are accelerometer and gyroscope noise, respectively.

[0064] In the INS mechanics choreography mode, the system state model for position, velocity, and attitude is as follows:

[0065] (15)

[0066] in, and These represent the rotational angular velocities of the e-frame (Geocentric coordinate system) relative to the i-frame (navigation coordinate system) and the n-frame relative to the e-frame, respectively. Let be the rotational angular velocity of the n-system relative to the i-system in the n-system. Let gravitational acceleration be in the n-system. and These represent the specific force error and angular rate error output by the IMU, respectively. For the gravity error in the n-system, construct the corresponding system state transition matrix based on the above system state model.

[0067] Preferably, in the multi-source information adaptive processing module, the observation and constraint information used in different modes are different. Therefore, a comprehensive observation model is constructed to unify multiple positioning methods, as follows:

[0068] (16)

[0069] (17)

[0070] in, and These are position and velocity observations, respectively. For magnetometer observations, For barometer elevation observation, For zero-velocity correction observations, For heading-constrained observations, The variance of each observation is... and These are the variances of the position and velocity observations, respectively. and These represent the variances of the observations from the magnetometer and barometer, respectively. and These are the variances of the zero-velocity correction and heading constraint observations, respectively. To adapt the observation vector adjustment coefficient matrix for different observation times, the adaptive adjustment coefficient matrix for the observation vector is determined based on user status and multi-sensor availability, and its form is as follows:

[0071] (18)

[0072] (19)

[0073] in, For adaptive adjustment coefficients, when the corresponding observations can be used for adaptive processing of multi-source information, It is 1 if it is true, otherwise it is 0.

[0074] Preferably, in the multi-source information adaptive processing module, the state prediction and observation correction process is performed based on the constructed system state model (formula (13)) and the comprehensive observation model (formula (16)):

[0075] (20)

[0076] ;(twenty one)

[0077] ;(twenty two)

[0078] ;(twenty three)

[0079] ;(twenty four)

[0080] Where the superscript T denotes the transpose of the matrix, and Let these represent the gain matrix and the identity matrix, respectively. First, a time update process is performed, utilizing... The state vector information and state transition matrix at each time step predict the current epoch. State vector at time step and the corresponding covariance matrix Then, an observation update process is performed, updating the state variables based on the observation information of the current epoch.

[0081] like Figure 2 As shown, the present invention also provides a multi-source adaptive fusion positioning method for IoT terminals, comprising the following steps:

[0082] Step 1: Set up multiple sensor modules, including: GNSS module, inertial navigation module, barometer and magnetometer module, to acquire different types of positioning and auxiliary data.

[0083] Step 2: Data preprocessing and user status identification: Each sensor module collects and performs preliminary data processing according to its own function. For example, the GNSS module acquires satellite navigation and positioning information, the barometer and magnetometer modules acquire barometric and magnetic data, the inertial navigation module acquires IMU data, and the user status identification module uses IMU data for smoothing and status identification. Based on the identification results, the INS mechanical arrangement and PDR process are performed.

[0084] Step 3, Multi-source information adaptive processing: Input the output data of each sensor module into the multi-source information adaptive processing module, and realize adaptive switching when fusing information from different sensors by constructing a unified observation model and block-based state coefficients (adaptive adjustment coefficient matrix).

[0085] Step 4, Kalman Filter Fusion: The Kalman filter algorithm is used to fuse multi-source data. The observations are dynamically adjusted according to the user status and the availability of multiple sensors to avoid redundant calculations caused by switching between different positioning algorithms and improve computational efficiency.

[0086] The method of this invention supports multiple combined positioning algorithms, and can flexibly select GNSS / INS combination, GNSS / PDR combination, and GNSS / PDR / magnetometer / barometer combination positioning algorithms according to different scenarios and motion states, providing users with accurate positioning services. Among them, GNSS / INS combination is suitable for most vehicle-mounted scenarios, such as when IoT devices are fixed to a carrier while riding or in a vehicle. GNSS / PDR combination is mainly suitable for pedestrian scenarios, such as when pedestrians are holding IoT devices while stationary, walking, or running. In addition, if the magnetometer and barometer sensors support it, the GNSS / PDR / magnetometer / barometer combined positioning algorithm can be further used in pedestrian scenarios.

[0087] Finally, the system outputs the final location result, providing users with accurate location services and an intelligent navigation experience.

[0088] Preferably, in step 2, based on the differences in time-domain characteristics of IMU accelerometer and gyroscope data under pedestrian and vehicle states, multiple feature values ​​such as accelerometer peak ratio, variance, autocorrelation degree, and gyroscope commutation frequency are extracted as identification criteria to construct a user state identification model based on multi-dimensional IMU features.

[0089] Specifically, in step 2, moving average filtering is used to smooth the noise in the IMU accelerometer and gyroscope data to reduce the impact of sensor noise. The peak-to-peak ratio and variance features are then extracted from the smoothed IMU accelerometer data, as follows:

[0090] ;

[0091] ;

[0092] in, The peak ratio characteristic value of the accelerometer. Indicates the IMU epoch. The total number of samples taken in the feature value extraction window. Indicates the first Z-axis acceleration of each epoch and These represent the maximum and average values ​​of the Z-axis acceleration, respectively. For the accelerometer variance eigenvalues, and These represent accelerations along the X and Y axes, respectively. and These are the average values ​​of the X-axis and Y-axis accelerations within the window epoch, respectively.

[0093] Specifically, in step 2, the autocorrelation of the accelerometer data is calculated to identify periodic characteristics during walking. The autocorrelation index is obtained as follows:

[0094] ;

[0095] ;

[0096] ;

[0097] in, This represents the hysteresis search time within the desired step frequency range, determined by combining the IMU sampling rate and pedestrian movement time. The value of , For accelerometer autocorrelation parameters, For delay The accelerometer autocorrelation function value at time 1. This is the accelerometer's self-power value. The resultant acceleration of the three-axis accelerometer. and This also represents the total number of IMU epochs and the feature extraction window IMU epochs.

[0098] Specifically, in step 2, the user's motion state is determined using gyroscope motion characteristics as an aid, and the angular velocity commutation frequency and angular velocity fluctuation variance are extracted as feature values, defined as follows:

[0099] ;

[0100] ;

[0101] in, This is the characteristic value of the gyroscope's commutation frequency. The window duration. This indicates that the value is 1 when the direction of the gyroscope's Z-axis angular velocity changes (i.e., changes from positive to negative). Indicates the first The angular velocity values ​​of each sampling point in the Z-axis direction; These are the eigenvalues ​​of the gyroscope's variance. The average Z-axis angular velocity within the window and This also represents the total number of IMU epochs and the feature extraction window IMU epochs.

[0102] Preferably, in step 2, by combining the aforementioned IMU motion time-domain features with a feature confidence adjustment mechanism and a normalization function, a state recognition model based on multi-dimensional IMU features is constructed, as follows:

[0103] ;

[0104] This involves using a sliding window process on the IMU data. For the first The state recognition score of each window For feature value index, The total number of eigenvalues ​​used. For the first The static weights of each feature represent the fundamental importance of that feature in the overall discrimination. The normalization function output for the j-th feature standardizes different feature values ​​to the interval [0,1]. The feature confidence factor and feature normalization function are as follows:

[0105] ;

[0106] Among them, for the feature normalization function , This is the original output of the j-th feature. and These are the minimum and maximum values ​​of the j-th feature in the empirical data, respectively. By standardizing the original features to the [0,1] interval, they are used for unified scoring across features.

[0107] Preferably, in step 3, a multi-source fusion extended Kalman filter model is constructed, including a state-space model and a multi-source data integrated observation model. In actual data processing, its linearized model is as follows:

[0108] ;

[0109] ;

[0110] Where the subscripts k and k-1 represent the epoch numbers, and the system state parameters and system noise at time k-1 are respectively... and At time k, the system state parameters are Observations are The observation state matrix is The observed noise is ; This represents the system state transition matrix from epoch k-1 to epoch k.

[0111] Preferably, in step 3, a system state model is constructed, using the inertial navigation system as the main filter, and the state parameter vectors related to the inertial navigation system are... Includes three-dimensional position, velocity, attitude, and zero bias of accelerometers and gyroscopes:

[0112] ;

[0113] In the formula, , and These are the position, velocity, and attitude error vectors, respectively. and These are the accelerometer zero bias and gyroscope zero bias error vectors, respectively.

[0114] Preferably, in step 3, considering the dynamic characteristics of INS information under different motion modes, different inertial navigation calculation methods are used in pedestrian mode and vehicle mode, including PDR mode and INS mechanical arrangement mode, and the inertial navigation system state model is switched according to different inertial navigation calculation methods:

[0115] ;

[0116] In the formula, This refers to the system state transition matrix corresponding to the position, velocity, and attitude parameters, which is switched according to different inertial navigation calculation methods. This is the state transition matrix corresponding to the zero bias error state parameters of the accelerometer and gyroscope. Given the state transition matrix between the position, velocity, and attitude state parameters and the zero bias error parameter, the system state model for position, velocity, and attitude, calculated using the PDR method, can be described as follows:

[0117] ;

[0118] In the formula, , and In order to be in The derivatives of position error, velocity error, and attitude error in the navigation coordinate system. For the projection of the force in the n-system, For the attitude angle error based on the PSI angle model, Let be the rotation matrix from frame b (the carrier coordinate system) to frame n. and These are accelerometer and gyroscope noise, respectively.

[0119] In the INS mechanics choreography mode, the system state model for position, velocity, and attitude is as follows:

[0120] ;

[0121] in, and These represent the rotational angular velocities of the e-frame (Geocentric coordinate system) relative to the i-frame (inertial coordinate system) and the n-frame relative to the e-frame, respectively. Let be the rotational angular velocity of the n-system relative to the i-system in the n-system. Let gravitational acceleration be in the n-system. and These represent the specific force error and angular rate error output by the IMU, respectively. For the gravity error in the n-system, construct the corresponding system state transition matrix based on the above system state model.

[0122] Preferably, in step 3, a comprehensive observation model is constructed. Since the observations and constraint information used in different modes are different, a comprehensive observation model is constructed to unify multiple positioning methods.

[0123] ;

[0124] ;

[0125] in, and These are position and velocity observations, respectively. For magnetometer observations, For barometer elevation observation, For zero-velocity correction observations, For heading-constrained observations, The variance of each observation is... and These are the variances of the position and velocity observations, respectively. and These represent the variances of the observations from the magnetometer and barometer, respectively. and These are the variances of the zero-velocity correction and heading constraint observations, respectively. To adapt the observation vector adjustment coefficient matrix for different observation times, the adaptive adjustment coefficient matrix for the observation vector is determined based on user status and multi-sensor availability, and its form is as follows:

[0126] ;

[0127] ;

[0128] in, For adaptive adjustment coefficients, when the corresponding observations can be used for adaptive processing of multi-source information, It is 1 if it is true, otherwise it is 0.

[0129] Preferably, in step 4, state prediction and observation correction are performed based on the system state model and integrated observation model constructed in step 3:

[0130] ;

[0131] ;

[0132] ;

[0133] ;

[0134] ;

[0135] Where the superscript T denotes the transpose of the matrix, and Let these represent the gain matrix and the identity matrix, respectively. First, a time update process is performed, utilizing... The state vector information and state transition matrix at each time step predict the current epoch. State vector at time step and the corresponding covariance matrix Then, an observation update process is performed, updating the state variables based on the observation information of the current epoch.

[0136] In one embodiment, the process of processing combined navigation data collected by a group of IoT terminals is as follows:

[0137] 1. Acquire GNSS, IMU, barometer, and magnetometer data.

[0138] 2. Smooth and extract features from IMU data to identify user status:

[0139] (1) Preprocessing of the output data of accelerometer and gyroscope. The sampling rate of IMU data in IoT devices is usually 100Hz, so the window length is generally selected. The window size is 2s, the total number of windows N is 200, the window sliding step size is 20, and this window is used to perform moving filter smoothing and subsequent feature extraction.

[0140] (2) Calculate the peak accelerometer ratio in each window. acceleration variance Acceleration autocorrelation index gyroscope commutation frequency and gyroscope variance A total of 5 feature values ​​were extracted, including those for acceleration autocorrelation index. During calculation, the lag search time is usually set based on the normal walking frequency of pedestrians. This corresponds to a step frequency of approximately 1.6Hz~2.4Hz. (Settings...) ;

[0141] (3) Normalize the eigenvalues ​​based on the obtained eigenvalues, and set the corresponding minimum and maximum eigenvalues ​​for each type of eigenvalue. and Calibration values ​​can usually be obtained based on prior data collection and analysis, and then set. , These correspond to the minimum and maximum values ​​of the five types of features, respectively;

[0142] (4) Perform state recognition score calculation and state classification, usually setting the static weights of various feature values ​​as follows: Set the state classification threshold based on empirical values. Then it is in walking mode, if If it is in vehicle mode, then it is in vehicle mode; otherwise, it remains in the previous mode.

[0143] 3. Based on the user state identification results, construct a state prediction model and an observation correction model for Kalman filtering. If the user state identification result is walking mode, then... Figure 3 As shown, in walking mode, a GNSS / PDR combination mode is used, and a magnetometer and barometer are selected for auxiliary positioning according to user needs. If the GNSS / PDR / magnetometer / barometer combination is selected, the process is as follows:

[0144] (1) Based on the observation information of the gyroscope and the accelerator, perform INS mechanical choreography to obtain the INS mechanical choreography attitude results;

[0145] (2) Based on the observation information of the gyroscope, the meter and the magnetometer, PDR solution is performed, and Kalman filtering time update is performed according to the ground system state model under PDR mode to obtain the variance and covariance corresponding to the PDR result;

[0146] (3) The PDR result time is matched with the GNSS result time. If the match is unsuccessful, the next PDR cycle will begin. If the match is successful, the Kalman observation update process will begin and the integrated navigation result will be output.

[0147] (4) Based on the PDR and GNSS position results, the INS mechanical arrangement heading and the heading change rate of the magnetometer, and the elevation change rate of the barometer, construct the Kalman observation information vector and update the Kalman observation.

[0148] Specifically, if a GNSS / PDR / magnetometer / barometer combination is used, then the Kalman state prediction model should select the system state model for position, velocity, and attitude in PDR mode:

[0149] ;

[0150] in, The system state transition matrix is ​​the system state parameters including position, velocity, and attitude. Represents the identity matrix. Indicates the time interval between adjacent epochs. This represents the antisymmetric matrix of the specific force in the n-system.

[0151] During the Kalman observation correction process, the adaptive adjustment coefficient matrix of the observation vector in the integrated observation model is:

[0152] ;

[0153] Kalman time update and measurement update processes are performed based on the state prediction model and the observation correction model, respectively, to obtain the combined positioning results.

[0154] If the recognition result is vehicle mode, then as follows Figure 4 As shown, in cycling or vehicle-mounted mode, a GNSS / INS combined mode is used, with zero-speed constraint and heading constraint information as auxiliary means for motion state constraint. The process is as follows:

[0155] (1) Based on the observation information of the gyroscope and the table, perform INS mechanical arrangement and Kalman filtering time update to obtain the variance and covariance corresponding to the INS mechanical arrangement result;

[0156] (2) GNSS navigation and positioning results and their covariance acquisition;

[0157] (3) The INS navigation calculation result time is matched with the GNSS result time. If the matching is unsuccessful, the next INS mechanical arrangement cycle will be entered. If the matching is successful, the GNSS / INS loose combination mode will be entered.

[0158] (4) Based on the INS results (position and velocity) and GNSS results (position and velocity), construct the Kalman observation information vector. When the static motion state is met, zero-velocity and static heading constraint observation equations can be constructed to carry out the Kalman observation update process and output the combined navigation results.

[0159] Specifically, if a GNSS / INS combined approach is used, then the Kalman state prediction model should select the system state model for position, velocity, and attitude under the INS mechanics calculation mode:

[0160] ;

[0161] ;

[0162] in, To represent an antisymmetric matrix, such as Represents the angular velocity of the n-frame relative to the e-frame. antisymmetric matrix, Represents the identity matrix. Indicates the time interval between adjacent epochs. Represents gravitational acceleration. This represents the transformation matrix between the navigation coordinate system and the geocentric coordinate system. Let be the rotational angular velocity of the e-frame (geocentric coordinate system) relative to the i-frame (inertial coordinate system) in the n-frame. Let n be the rotational angular velocity of the n-system relative to the i-system. and These are the radius of curvature of the meridian and the radius of curvature of the zonal circle, respectively. This represents the elevation value.

[0163] Finally, the adaptive adjustment coefficient matrix of the observation vector of the integrated observation model in the Kalman observation correction process is obtained:

[0164] ;

[0165] ;

[0166] in, For adaptive adjustment coefficients, when the corresponding observations can be used for adaptive processing of multi-source information, It is 1 if it is true, otherwise it is 0.

[0167] 4. Perform Kalman time update and measurement update processes based on the state prediction model and observation correction model respectively to obtain the combined positioning results.

[0168] Based on the parameters and procedures in the above embodiments, a set of data under vehicle motion conditions is processed, and a comparison of the positioning results of the present invention and the traditional vehicle-mounted combined positioning mode is shown in the figure below. Figure 5 As shown, the solid line represents the result under the method of this invention, and the dashed line represents the result under the traditional vehicle-mounted integrated positioning method. The position is transformed to the northeast-northeast coordinate system, with the horizontal and vertical coordinates representing the coordinates in the east and north directions, respectively. The results show that the result under this invention can correctly identify the vehicle's motion state and has high consistency with the result under the traditional vehicle-mounted integrated positioning mode. Furthermore, the addition of motion constraints improves the positioning effect.

[0169] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described multi-source adaptive fusion positioning method for an Internet of Things terminal.

[0170] The present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described multi-source adaptive fusion positioning method for an Internet of Things terminal.

[0171] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0172] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0173] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0174] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0175] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0176] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A multi-source adaptive fusion positioning system for Internet of Things (IoT) terminals, characterized in that, include: The GNSS module is used to acquire satellite navigation and positioning information; The inertial navigation module is used to acquire raw IMU data and provide INS mechanical orchestration and PDR processes; The user status recognition module is used to extract multi-dimensional motion features based on IMU data and output the pedestrian or vehicle status; the barometer and magnetometer modules are used to provide elevation and heading assistance information; the multi-source information adaptive processing module is used to receive data from the above modules, dynamically adjust the observation vector within the same observation model according to the user status and sensor availability, and output the fused positioning result through a single Kalman filter, realizing the seamless continuous operation of the positioning algorithm in different scenarios; IMU stands for Inertial Measurement Unit, INS stands for Inertial Navigation System, and PDR stands for Pedestrian Track Deduction. Different inertial navigation calculation methods are used in pedestrian mode and vehicle mode, including PDR mode and INS mechanical arrangement mode. The state model of the inertial navigation system is switched according to the different inertial navigation calculation methods. The system state transition matrix corresponding to the position, velocity, and attitude state parameters is switched according to different inertial navigation calculation methods; In walking mode, a GNSS / PDR combination mode is adopted, and magnetometers and barometers are selected for auxiliary positioning according to user needs. In cycling or vehicle mode, a GNSS / INS combined mode is adopted, and zero-speed constraint and heading constraint information are used as auxiliary to constrain motion state; The observation vector adaptive adjustment coefficient matrix is ​​determined based on user status and multi-sensor availability, and its form is as follows: （18） （19） in, For adaptive adjustment coefficients, when the corresponding observations can be used for adaptive processing of multi-source information, It is 1 if it is true, otherwise it is 0.

2. The IoT terminal multi-source adaptive fusion positioning system according to claim 1, characterized in that, The user status recognition module smooths the IMU data using a moving average filter, then extracts the peak acceleration ratio, variance, autocorrelation degree, and gyroscope commutation frequency characteristics, and determines the status through normalized weighted scoring.

3. The IoT terminal multi-source adaptive fusion positioning system according to claim 1, characterized in that, The multi-source information adaptive processing module primarily uses GNSS / PDR in pedestrian mode, combined with constraints from magnetometers and barometers. In vehicle mode, it primarily uses GNSS / INS, with zero-speed and heading constraints introduced.

4. The IoT terminal multi-source adaptive fusion positioning system according to claim 1, characterized in that, The elevation and heading information output by the barometer and magnetometer modules are selectively introduced into the filter through an adaptive adjustment coefficient matrix. The coefficient is set to 1 when the corresponding sensor is available and to 0 when it is unavailable.

5. A multi-source adaptive fusion positioning method for IoT terminals, characterized in that, Includes the following steps: Step 1: Synchronously acquire raw data from GNSS, IMU, barometer, and magnetometer; Step 2: After smoothing the IMU data, extract multi-dimensional temporal features and determine the status of pedestrians or vehicles in real time based on the feature scores. Step 3: Under the same Kalman filter framework, dynamically adjust the observation vector based on the state results of Step 2 and sensor availability, and integrate GNSS, INS / PDR, barometer and magnetometer information. Step 4: Switch the state model of the inertial navigation system according to the different inertial navigation calculation methods in pedestrian mode or vehicle mode, and introduce the corresponding motion constraints. Output the continuous positioning results directly through one-step Kalman time update and observation update. Different inertial navigation calculation methods are used in pedestrian mode and vehicle mode, including PDR mode and INS mechanical arrangement mode. The state model of the inertial navigation system is switched according to the different inertial navigation calculation methods. The system state transition matrix corresponding to the position, velocity, and attitude state parameters is switched according to different inertial navigation calculation methods; In walking mode, a GNSS / PDR combination mode is adopted, and magnetometers and barometers are selected for auxiliary positioning according to user needs. In cycling or vehicle mode, a GNSS / INS combined mode is adopted, and zero-speed constraint and heading constraint information are used as auxiliary to constrain motion state; The observation vector adaptive adjustment coefficient matrix is ​​determined based on user status and multi-sensor availability, and its form is as follows: （18） （19） in, For adaptive adjustment coefficients, when the corresponding observations can be used for adaptive processing of multi-source information, It is 1 if it is true, otherwise it is 0.

6. The multi-source adaptive fusion positioning method for IoT terminals according to claim 5, characterized in that, In step 2, feature extraction is completed within a sliding window of 200 sampling points. The obtained features are normalized and weighted to obtain a state score, which is then compared with a set threshold to lock the current motion state.

7. The multi-source adaptive fusion positioning method for IoT terminals according to claim 6, characterized in that, Step 3 achieves dynamic adjustment by constructing an adaptive adjustment coefficient matrix for the observation vector. The matrix elements are set to 1 or 0 based on sensor availability, and the corresponding observation is determined in real time whether it participates in the Kalman filter update.

8. The multi-source adaptive fusion positioning method for IoT terminals according to claim 7, characterized in that, When step 2 determines that it is a pedestrian mode, step 4 first aligns the PDR results with the GNSS position in time, then incorporates the magnetometer heading change rate and barometer elevation change rate into the observation information, and then performs Kalman observation update.

9. The multi-source adaptive fusion positioning method for IoT terminals according to claim 8, characterized in that, When step 2 determines that it is in vehicle mode, step 4 automatically introduces zero-speed correction and static heading constraint in a stationary state based on the loose combination of GNSS / INS, and completes the observation update and outputs the integrated navigation result through the same Kalman filter.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the multi-source adaptive fusion positioning method for an Internet of Things terminal as described in any one of claims 5 to 9.

11. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-source adaptive fusion positioning method for an Internet of Things terminal as described in any one of claims 5 to 9.

Images