Accurate positioning fusion redundancy method, system and device and storage medium

Abstract

The invention discloses an accurate positioning fusion redundancy method, system and device and a storage medium, and the method comprises the steps: collecting the state information of each mechanism through a sensor, obtaining an observation state, and enabling the state to comprise the position and speed of the mechanism; carrying out RLS online self-calibration on the observation state of the redundant sensor; according to the fusion state of the previous moment and system input, calculating a prediction state of the current moment; deriving a theoretical observation value of each sensor in the state according to the prediction state; calculating a residual error between the observation state of the sensor after the RLS online self-calibration and a theoretical observation value; when the residual error is greater than a threshold value, eliminating an abnormal observation state of the sensor; dynamically adjusting the fusion weight according to the confidence coefficient and the residual error of the observation state of each sensor; and fusing the prediction state and the observation state of the sensor based on the fusion weight to obtain a fusion state at the current moment. According to the method, the positioning precision is remarkably improved, and the system robustness and reliability are enhanced.

Description

Technical Field

[0001] This invention relates to crane positioning technology, and more specifically to a precise positioning fusion redundancy method, system, device, and storage medium. Background Technology

[0002] Port cranes are widely used in port settings such as container terminals and bulk cargo terminals, undertaking critical tasks such as loading, unloading, and transshipment of bulk cargo. To ensure operational efficiency and safety, real-time and accurate status monitoring and position measurement of each core motion mechanism of the crane (such as hoisting, luffing, slewing, and traveling) are required. Currently, the positioning and attitude perception of port cranes typically rely on multiple sensor units, such as absolute encoders, incremental encoders, laser rangefinders, RTK-GNSS, inclinometers, and RFID. Most systems often adopt an independent measurement mode with a single sensor and single loop, or only perform simple serial correction, lacking the ability to deeply couple and jointly optimize multi-sensor data in real time at the algorithm level.

[0003] Chinese patent CN113375666B proposes a sensor fusion positioning system and method for autonomous vehicles. The core design includes: 1) Multi-module parallel computation: The main navigation module and redundant navigation modules receive multi-source sensor data (such as IMU, GNSS, and LiDAR) from their respective configurations, independently run different sensor fusion algorithms (such as Kalman filtering and factor graph optimization), and each outputs a complete set of vehicle state information. 2) Fault self-diagnosis: A fault diagnosis module is set up to monitor the consistency and smoothness between the output data of two or more navigation modules in real time, and determine whether the module operation is abnormal through methods such as state residual analysis and rate of change analysis. 3) Output switching and redundancy backup: Through an output selection module, the most reliable navigation module result is dynamically selected as the final vehicle positioning output based on the diagnostic results. Once the main module (e.g., relying on GNSS) fails due to signal failure or environmental obstruction (such as tunnels), the redundant module (e.g., relying on LiDAR and odometer) can immediately take over the output, ensuring continuous and stable positioning of the autonomous vehicle.

[0004] The existing technical solutions have the following drawbacks: (1) Lack of deep data fusion leads to information waste and accuracy bottlenecks: In existing technologies, although the main navigation module and the redundant navigation module each access multiple sensor data and run in parallel, they are independent at the physical and algorithmic levels. Their outputs are only used for later availability switching or consistency diagnosis, rather than for mutual correction and gain. Therefore, the measurement data of key high-precision sensors (such as absolute encoders) are not used for online calibration of the drift and system errors of other sensors in the redundant module, and vice versa. This results in the system having to select a single path as the final output from multiple sub-modules, failing to form true information complementarity and accuracy superposition. The upper limit of positioning accuracy is locked at the optimal level of a single path, making it difficult to achieve a fusion gain effect of 1+1>2.

[0005] (2) Lack of integrated sensor coupling and cross-calibration mechanism for multiple mechanisms: Existing autonomous vehicle solutions focus on vehicle positioning and emphasize global pose calculation. However, port cranes usually involve multiple independent motion mechanisms (such as hoisting, luffing, slewing, and traveling). The motion modes and sensor types of each mechanism are diverse. If only a single modular parallel calculation is relied upon, it is difficult to fully cross-check and complementarily correct the coupling error and sensor drift between the mechanisms. Summary of the Invention

[0006] This invention addresses the problems of low accuracy and poor robustness in existing port crane positioning systems by providing a precise positioning fusion redundancy method, system, equipment, and storage medium. Redundant sensors are configured for each motion mechanism of the crane (such as hoisting, luffing, rotation, and traveling). Based on the Jacobian matrix and recursive least squares method, online self-calibration and weighted fusion of positioning data are performed, effectively eliminating system errors and random noise, and real-time filtering of abnormal observations. This significantly improves the accuracy, robustness, and adaptability of the positioning system, providing reliable technical support for the automated and intelligent operation of port cranes.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A precise positioning and redundancy fusion method is used for the positioning of various mechanisms of a crane, and the following steps are performed in each iteration cycle: The sensors collect the state information of each mechanism to obtain the observed state, which includes the position and velocity of the mechanism. Perform online self-calibration of the observation status of redundant sensors using RLS; Based on the fusion state and system input of the previous moment, the predicted state of the current moment is calculated; The theoretical observation values ​​of each sensor under the predicted state are derived. The residual between the observation status of the sensor after RLS online self-calibration and the theoretical observation value is calculated; when the residual is greater than the threshold, the abnormal observation status of the sensor is removed. The fusion weights are dynamically adjusted based on the confidence level and residuals of the observations from each sensor. The fused state at the current moment is obtained by fusing the predicted state and the sensor's observation state based on the fusion weight.

[0008] To optimize the above technical solution, the specific measures also include: Furthermore, the RLS online self-calibration specifically includes: The formula for RLS online self-calibration is as follows:

[0009] In the formula, This is the observation status of the sensor after RLS online self-calibration. X i ( k ) is the first i A redundant sensor k The original observation state at that moment, a i ( k This is a proportionality coefficient reflecting the sensor's sensitivity error; b i ( k To reflect the bias amount of zero-point drift, before calibration using the RLS online self-calibration formula in each iteration, the proportional coefficient and the bias amount are updated. The update process is as follows: Define parameter vector With regression vector :

[0010] For parameter vectors Perform the following iterative updates:

[0011] In the formula, K i (k) is the gain coefficient. P i ( k Let be the parameter covariance matrix at time k. P i ( k -1) is the parameter covariance matrix at time k-1. X ref ( k (This refers to the actual output of the reference sensor.) λ It is a forgetting factor.

[0012] Furthermore, the step of calculating the predicted state at the current moment based on the fusion state and system input at the previous moment specifically involves:

[0013] In the formula, It is the predicted state at the current moment. , This is the predicted location of the institution at the current moment. This is the predicted speed of the institution at the current moment. t It is the sampling period. u ( k -1) is the system input at the previous moment, i.e., the system acceleration. It represents the fusion state from the previous moment. , It is the location of the institution that was integrated at the previous moment. It is the speed of the institutions that merged at the previous moment.

[0014] Furthermore, the derivation of the theoretical observation values ​​of each sensor under the predicted state specifically involves:

[0015] In the formula, Z pred , i ( k ) is the derivation of the first i Theoretical observations from each sensor, h i (·) is the first i The observation function of each sensor, It is the predicted state at the current moment; Calculate the Jacobian matrix, which will be used to calculate the Kalman gain in the subsequent fusion step; The formula for calculating the Jacobian matrix is ​​as follows:

[0016] In the formula, H i ( k ) is the Jacobian matrix, h i For the first i The observation function of each sensor.

[0017] Furthermore, the residual is calculated as follows:

[0018] In the formula, e i ( k ) represents the residual. X ′ i ( kThis represents the sensor's observation status after RLS online self-calibration. Z pred , i ( k ) is the derivation of the first i Theoretical observations from each sensor; The conditions for excluding abnormal sensor observation states are as follows:

[0019] In the formula, e th The residual threshold, W i ( k ) represents the sensor weight, which is set to zero in case of an anomaly.

[0020] Furthermore, the specific steps of dynamically adjusting the fusion weights based on the confidence level and residuals of each sensor's observation state are as follows:

[0021] In the formula, W i ( k ) is the first i The weight of each sensor, For the first i The noise variance of each sensor at time k. n This represents the number of sensors.

[0022] Furthermore, the specific details of the fusion prediction state based on the fusion weight and the sensor's observation state are as follows: The Kalman gain is calculated using the Jacobian matrix, as shown in the following formula:

[0023] In the formula, K ( k ) represents the Kalman gain. P pred ( k ) represents the covariance matrix of the predicted state; H ( k ) is the Jacobian matrix; R ( k The observation noise covariance matrix is ​​denoted by T, where T denotes transpose. The predicted state and the sensor's observed state are fused based on Kalman gain and fusion weights for state updating, as follows:

[0024] In the formula, The current state of fusion. It is the predicted state at the current moment.K ( k ) represents the Kalman gain. W i ( k ) is the first i The weight of each sensor, e i ( k ) represents the residual. W i ( k )· e i ( k This represents the weighted contribution of each sensor residual to the state update; The covariance is updated as follows:

[0025] In the formula, P ( k () represents the updated covariance matrix. I It is the identity matrix. K ( k ) represents the Kalman gain. H ( k ) is the Jacobian matrix, P pred ( k ) represents the covariance matrix of the predicted state; Updated covariance matrix P ( k This serves as the initial uncertainty input for the state prediction phase of the next iteration cycle.

[0026] This invention also proposes a precise positioning fusion redundancy system, comprising: Sensors are used to collect state information of each mechanism to obtain the observed state, which includes the position and velocity of the mechanism. The data preprocessing module is used to filter, convert units, align timestamps, and initially remove outliers from the data collected by the sensors. The fusion calibration and positioning calculation module is used to perform online RLS self-calibration on the observation states of redundant sensors; to calculate the predicted state at the current time based on the fusion state and system input at the previous time step; to derive the theoretical observation values ​​of each sensor in this state based on the predicted state; to calculate the residual between the observation states of the sensors after online RLS self-calibration and the theoretical observation values; to remove the observation states of abnormal sensors when the residual is greater than a threshold; to dynamically adjust the fusion weights based on the confidence level and residuals of each sensor's observation state; and to obtain the fusion state at the current time step by fusing the predicted state and the sensor's observation state based on the fusion weights. The output module is used to extract the fused position from the fused state and output the real-time position estimate.

[0027] The present invention also proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the precise positioning fusion redundancy method as described above.

[0028] A computer-readable storage medium storing a computer program that causes a computer to perform the precise positioning and redundancy fusion method as described above.

[0029] The beneficial effects of this invention are: (1) Significantly improves positioning accuracy: This invention achieves deep coupling and online self-calibration of various sensor data through a multi-sensor fusion algorithm based on Jacobian matrix and recursive least squares (RLS). Unlike existing single-sensor or simple calibration methods, this method can dynamically correct proportional error and bias error during the solution process, effectively suppress random noise, and achieve full complementarity of information and accuracy superposition, thereby obtaining high-precision positioning results far exceeding those of a single path.

[0030] (2) Enhancing System Robustness and Reliability: This invention equips key mechanisms such as hoisting, luffing, rotation, and traveling with redundant sensors. Through residual detection, anomaly removal, and dynamic weight allocation mechanisms, it ensures that when some sensors malfunction, produce abnormal outputs, or experience performance degradation, the system can immediately identify and reduce their weights or remove the observed values, continuing to rely on the remaining sensors to complete the positioning calculation. This proactive fault-tolerant and adaptive mechanism significantly improves the stability and reliability of the system under complex port conditions.

[0031] (3) Adaptability and scalability: This invention introduces a forgetting factor and a dynamic weight adjustment strategy, enabling the system to track sensor performance changes during long-term operation and achieve online self-correction and continuous optimization of the positioning algorithm. At the same time, the algorithm structure has good versatility and scalability, and can be flexibly adapted to different types or numbers of sensors, making it easy to promote its application in various port equipment and complex operating environments. Attached Figure Description

[0032] Figure 1 This is a flowchart of the precise positioning and redundancy fusion method proposed in this invention. Detailed Implementation

[0033] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0034] Example 1 This invention proposes a precise positioning fusion redundancy method for positioning various mechanisms of a crane. Each mechanism (lifting, luffing, slewing, and traveling) is equipped with an independent redundant sensor group and independently runs a complete fusion calibration algorithm. This not only reduces the complexity of the algorithm but also improves the modularity and reliability of the system. In this way, high-precision real-time position or status information (such as lifting height, working radius, slewing angle, and traveling position) of each mechanism can be obtained.

[0035] Lifting mechanism: adopts a combination of absolute encoder and incremental encoder.

[0036] Amplitude adjustment mechanism: A combination of absolute encoder and inclinometer is used.

[0037] Rotation mechanism: It adopts a combination of absolute encoder and QR code camera to accurately measure the rotation angle of the door operator.

[0038] The traveling mechanism employs a combination of absolute encoders and RFID tags. While absolute encoders provide continuous travel position, they are susceptible to cumulative errors due to wheel slippage. RFID tags laid along the track provide discrete, absolutely accurate position reference points for periodically correcting and resetting the encoder's cumulative errors.

[0039] The overall flow of the method in this example is as follows: Figure 1 As shown, the following steps are performed in each iteration cycle: The sensors collect the state information of each mechanism to obtain the observed state, which includes the position and velocity of the mechanism. The error model of the sensor is:

[0040] In the formula, X ref ( k The reference sensor's actual output is used (select a sensor with high accuracy or excellent stability). X i ( k ) is the first i A redundant sensor at time k The original output; a i ( k ) is the proportionality coefficient (reflecting the sensor sensitivity error); b i ( k ) is the bias (reflecting zero drift).

[0041] Define the state vector as a two-dimensional column vector containing position and velocity:

[0042] In the formula, p ( k ) represents the positional variables of the mechanism (such as lifting height, working radius, slewing angle, and traveling position); v ( k ) represents the velocity variable of the mechanism.

[0043] The observation status of redundant sensors is subjected to online self-calibration using Recursive Least Squares (RLS). The formula for online RLS self-calibration is as follows:

[0044] In the formula, This is the observation status of the sensor after RLS online self-calibration. X i ( k ) is the first i A redundant sensor k The original observation state at that moment, a i ( k This is a proportionality coefficient reflecting the sensor's sensitivity error; b i ( k To reflect the bias amount of zero-point drift, before calibration using the RLS online self-calibration formula in each iteration, the proportional coefficient and the bias amount are updated. The update process is as follows: Define parameter vector With regression vector :

[0045] For parameter vectors Perform the following iterative updates:

[0046] In the formula, K i (k) is the gain coefficient, which controls the magnitude of parameter updates from new observations. P i ( k Let be the parameter covariance matrix at time k, representing the uncertainty of the parameter estimation; P i ( k -1) is the parameter covariance matrix at time k-1. X ref ( k (This refers to the actual output of the reference sensor.) λThe forgetting factor, ranging from (0,1], is used to adjust the impact of historical data on the current parameter estimation. A smaller value allows the algorithm to adapt more quickly to changes in sensor characteristics.

[0047] Based on the fused state and system input from the previous moment, the predicted state for the current moment is calculated. The core idea is to use the kinematic equations of the mechanism to describe the evolution of the state over time, enabling the system to maintain continuous and stable state estimation even when sensor data is temporarily missing or affected by noise. This invention uses a uniformly accelerated motion model to model the system. During the sampling period... t Within the context, the discrete-time evolution of the state is as follows: Current location p ( k )= p ( k -1)+ v ( k -1)· t+a ( k -1)· t 2 / 2 Current speed v ( k )= v ( k -1)+ a ( k -1)· t The state prediction model is represented in matrix form:

[0048] In the formula, It is the predicted state at the current moment. , This is the predicted location of the institution at the current moment. This is the predicted speed of the institution at the current moment. t It is the sampling period. u ( k -1) is the system input at the previous moment, i.e., the system acceleration. It represents the fusion state from the previous moment. , It is the location of the institution that was integrated at the previous moment. It is the speed of the institutions that merged at the previous moment.

[0049] The theoretical observation values ​​of each sensor under the predicted state are derived; specifically:

[0050] In the formula, Z pred , i ( k ) is the derivation of the first i Theoretical observations from each sensor, h i (·) is the first i The observation function of each sensor, It is the predicted state at the current moment; For nonlinear observations, linearization is required at the predicted state, and the Jacobian matrix is ​​calculated for use in the subsequent fusion step to calculate the Kalman gain. The formula for calculating the Jacobian matrix is ​​as follows:

[0051] In the formula, H i ( k Let be the Jacobian matrix, describing the sensitivity of state changes to observations. If H i ( k If a certain component of the state variable is relatively large, it indicates that the state variable is sensitive to the output of the sensor, and the algorithm will give it a higher weight. h i For the first i The observation function of each sensor describes the mapping relationship between the system state and the sensor output.

[0052] The residuals are calculated between the observation status of the sensors after RLS online self-calibration and the theoretical observation values. When the residuals are greater than a threshold, the observation status of abnormal sensors is removed. This step identifies abnormal sensor data and suppresses its impact on the fusion results by quantitatively analyzing the difference between the predicted and actual observations, thus ensuring the robustness and reliability of the fusion process.

[0053] The residuals are calculated as follows:

[0054] In the formula, e i ( k ) represents the residual. X ′ i ( k This represents the sensor's observation status after RLS online self-calibration. Z pred , i ( k ) is the derivation of the first i Theoretical observations from each sensor; The conditions for excluding abnormal sensor observation states are as follows:

[0055] In the formula, e th The residual threshold, W i ( k The weights are set to zero when data is deemed abnormal. For data identified as abnormal, the system will reset their weights to zero, preventing them from participating in subsequent weighted fusion calculations, thus achieving removal at the algorithm level.

[0056] The fusion weights are dynamically adjusted based on the confidence level and residuals of each sensor's observation status; specifically:

[0057] In the formula, W i ( k ) is the first i The weight of each sensor, For the first i The noise variance of each sensor at time k. n This represents the number of sensors.

[0058] The fused state at the current moment is obtained by fusing the predicted state and the sensor's observed state based on the fusion weights. Specifically: The Kalman gain is calculated using the Jacobian matrix, as shown in the following formula:

[0059] In the formula, K ( k The Kalman gain determines the fusion ratio of predictions and observations. P pred ( k ) represents the covariance matrix of the predicted state; H ( k ) is the Jacobian matrix; R ( k Let T be the observation noise covariance matrix, and the superscript T denotes transpose. When the observation noise is low and the prediction variance is high, the system tends to trust the observations more; otherwise, it relies more on the model prediction results. This mechanism ensures an adaptive balance of the fusion results under different sensor reliability levels.

[0060] The predicted state and the sensor's observed state are fused based on Kalman gain and fusion weights for state updating, as follows:

[0061] In the formula, The current state of fusion. It is the predicted state at the current moment. K ( k ) represents the Kalman gain. W i ( k ) is the first i The weight of each sensor, e i ( k ) represents the residual. W i ( k )· e i ( k The weighted contribution of each sensor residual to the state update is used; by correcting the predicted state through weighted residuals, the system can achieve dynamic balance among multi-source observation information, effectively suppress the impact of single sensor anomalies on the positioning results, thereby improving the overall robustness and stability of the system.

[0062] The covariance is updated as follows:

[0063] In the formula, P ( k () represents the updated covariance matrix. I It is the identity matrix. K ( k ) represents the Kalman gain. H ( k ) is the Jacobian matrix, P pred ( k ) represents the covariance matrix of the predicted state; Updated covariance matrix P ( k This serves as the initial uncertainty input for the state prediction stage of the next iteration cycle, thus forming a recursive closed-loop estimation process. This recursive characteristic enables the system to continuously correct and optimize the state estimation, achieving real-time, high-precision dynamic positioning.

[0064] Example 2 This invention proposes a precise positioning fusion redundancy system corresponding to the method in Embodiment 1, comprising: Sensors are used to collect state information of each mechanism to obtain the observed state, which includes the position and velocity of the mechanism. The data preprocessing module is used to filter, convert units, align timestamps, and initially remove outliers from the data collected by the sensors. The fusion calibration and positioning calculation module is used to perform online RLS self-calibration on the observation states of redundant sensors; to calculate the predicted state at the current time based on the fusion state and system input at the previous time step; to derive the theoretical observation values ​​of each sensor in this state based on the predicted state; to calculate the residual between the observation states of the sensors after online RLS self-calibration and the theoretical observation values; to remove the observation states of abnormal sensors when the residual is greater than a threshold; to dynamically adjust the fusion weights based on the confidence level and residuals of each sensor's observation state; and to obtain the fusion state at the current time step by fusing the predicted state and the sensor's observation state based on the fusion weights. The output module is used to extract the fused position from the fused state and output the real-time position estimate.

[0065] The implementation methods of each module and its function in the system are completely consistent with the steps of the method in Implementation Example 1, so they will not be repeated here.

[0066] Example 3 The present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the precise positioning fusion redundancy method as described in Embodiment 1.

[0067] Example 4 The present invention proposes a computer-readable storage medium storing a computer program that causes a computer to execute the precise positioning and redundancy fusion method as described in Embodiment 1.

[0068] In the embodiments disclosed in this application, a computer storage medium may be a tangible medium that may contain or store programs for use by or in conjunction with an instruction execution system, apparatus, or device. The computer storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of computer storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, and portable compact disc read-only memory (CD). ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0069] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0070] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A precise positioning and redundancy fusion method for positioning various mechanisms of a crane, characterized in that, Perform the following steps in each iteration cycle: The sensors collect the state information of each mechanism to obtain the observed state, which includes the position and velocity of the mechanism. Perform online self-calibration of the observation status of redundant sensors using RLS; Based on the fusion state and system input of the previous moment, the predicted state of the current moment is calculated; The theoretical observation values ​​of each sensor under the predicted state are derived. The residual between the observation status of the sensor after RLS online self-calibration and the theoretical observation value is calculated; when the residual is greater than the threshold, the abnormal observation status of the sensor is removed. The fusion weights are dynamically adjusted based on the confidence level and residuals of the observations from each sensor. The fused state at the current moment is obtained by fusing the predicted state and the sensor's observation state based on the fusion weight.

2. The precise positioning and redundancy fusion method as described in claim 1, characterized in that, The RLS online self-calibration specifically refers to: The formula for RLS online self-calibration is as follows: In the formula, This is the observation status of the sensor after RLS online self-calibration. X i ( k ) is the first i A redundant sensor k The original observation state at that moment, a i ( k This is a proportionality coefficient reflecting the sensor's sensitivity error; b i ( k To reflect the bias amount of zero-point drift, before calibration using the RLS online self-calibration formula in each iteration, the proportional coefficient and the bias amount are updated. The update process is as follows: Define parameter vector With regression vector : For parameter vectors Perform the following iterative updates: In the formula, K i (k) is the gain coefficient. P i ( k Let be the parameter covariance matrix at time k. P i ( k -1) is the parameter covariance matrix at time k-1. X ref ( k (This refers to the actual output of the reference sensor.) λ It is a forgetting factor.

3. The precise positioning and redundancy fusion method as described in claim 1, characterized in that, The specific steps for calculating the predicted state at the current moment based on the fusion state and system input at the previous moment are as follows: In the formula, It is the predicted state at the current moment. , This is the predicted location of the institution at the current moment. This is the predicted speed of the institution at the current moment. t It is the sampling period. u ( k -1) is the system input at the previous moment, i.e., the system acceleration. It represents the fusion state from the previous moment. , It is the location of the institution that was integrated at the previous moment. It is the speed of the institutions that merged at the previous moment.

4. The precise positioning and redundancy fusion method as described in claim 1, characterized in that, The specific steps for deriving the theoretical observation values ​​of each sensor under the predicted state are as follows: In the formula, Z pred , i ( k ) is the derivation of the first i Theoretical observations from each sensor, h i (·) is the first i The observation function of each sensor, It is the predicted state at the current moment; Calculate the Jacobian matrix, which will be used to calculate the Kalman gain in the subsequent fusion step; The formula for calculating the Jacobian matrix is ​​as follows: In the formula, H i ( k ) is the Jacobian matrix, h i For the first i The observation function of each sensor.

5. The precise positioning and redundancy fusion method as described in claim 1, characterized in that, The residual is calculated as follows: In the formula, e i ( k ) represents the residual. X ′ i ( k This represents the sensor's observation status after RLS online self-calibration. Z pred , i ( k ) is the derivation of the first i Theoretical observations from each sensor; The conditions for excluding abnormal sensor observation states are as follows: In the formula, e th The residual threshold, W i ( k ) represents the sensor weight, which is set to zero in case of an anomaly.

6. The precise positioning and redundancy fusion method as described in claim 1, characterized in that, The specific steps for dynamically adjusting the fusion weights based on the confidence level and residuals of each sensor's observation status are as follows: In the formula, W i ( k ) is the first i The weight of each sensor, For the first i The noise variance of each sensor at time k. n This represents the number of sensors.

7. The precise positioning and redundancy fusion method as described in claim 1, characterized in that, The specific details of the predicted state based on fusion weights and the sensor's observation state are as follows: The Kalman gain is calculated using the Jacobian matrix, as shown in the following formula: In the formula, K ( k ) represents the Kalman gain. P pred ( k () represents the covariance matrix of the predicted state; H ( k ) is the Jacobian matrix; R ( k The observation noise covariance matrix is ​​denoted by T, where T denotes transpose. The predicted state and the sensor's observed state are fused based on Kalman gain and fusion weights for state updating, as follows: In the formula, The current state of fusion. It is the predicted state at the current moment. K ( k ) represents the Kalman gain. W i ( k ) is the first i The weight of each sensor, e i ( k ) represents the residual. W i ( k )· e i ( k This represents the weighted contribution of each sensor residual to the state update; The covariance is updated as follows: In the formula, P ( k () represents the updated covariance matrix. I It is the identity matrix. K ( k ) represents the Kalman gain. H ( k ) is the Jacobian matrix, P pred ( k () represents the covariance matrix of the predicted state; Updated covariance matrix P ( k This serves as the initial uncertainty input for the state prediction phase of the next iteration cycle.

8. A precise positioning fusion redundancy system, characterized in that, include: Sensors are used to collect state information of each mechanism to obtain the observed state, which includes the position and velocity of the mechanism. The data preprocessing module is used to filter, convert units, align timestamps, and initially remove outliers from the data collected by the sensors. The fusion calibration and positioning calculation module is used to perform online RLS self-calibration on the observation status of redundant sensors; based on the fusion status and system input of the previous moment, it calculates the predicted status of the current moment; based on the predicted status, it derives the theoretical observation value of each sensor in this state; it calculates the residual between the observation status of the sensors after online RLS self-calibration and the theoretical observation value; when the residual is greater than the threshold, it removes the observation status of abnormal sensors. The fusion weights are dynamically adjusted based on the confidence level and residuals of the observations from each sensor. The fused state at the current moment is obtained by fusing the predicted state and the sensor's observation state based on the fusion weight; The output module is used to extract the fused position from the fused state and output the real-time position estimate.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the precise positioning fusion redundancy method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, The computer program causes the computer to execute the precise positioning and redundancy fusion method as described in any one of claims 1-7.

Images