A COA attitude solution method based on MEMS IMU assisted search

Abstract

The application provides a COA (Coati Optimization Algorithm) attitude solution method based on MEMS IMU auxiliary search, which comprises the following steps: collecting IMU and BDS double-antenna data in a city street where GNSS signals are weak; calculating a pitch angle and a yaw angle through MEMS IMU output data to determine an optimization algorithm search boundary; defining a fitness function to be optimized according to a GNSS carrier phase observation model and an attitude observation model; introducing an adaptive dynamic parameter adjustment mechanism to accelerate the convergence process of the algorithm and effectively avoid falling into a local optimal solution; aiming at the problem of too large jumping or slow convergence speed in the solution space, a step length adaptive adjustment mechanism is proposed to improve the efficiency and precision of the optimization process; and the improved COA is applied to attitude search and optimization within the space range determined by the inertial navigation system to solve the accurate attitude of the system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of GNSS high-precision positioning and attitude determination, and particularly relates to a COA attitude determination method based on MEMS IMU assisted search. BACKGROUND

[0002] In complex urban environments, GNSS (Global Navigation Satellite System) signals face numerous interferences and obstructions, especially in high-rise urban streets, narrow tunnels, and tree-dense rural paths. The signal propagation in these areas is severely tested, with significant signal quality degradation and even loss of lock, making it extremely difficult to rely solely on GNSS signals for positioning and attitude estimation. For example, in urban streets, it is almost inevitable to pass through densely built-up areas, small tunnels, and even underground parking lots without skylights. In these environments, satellite signal quality often deteriorates rapidly, and even complete loss of lock can occur, leading to a significant decrease in navigation accuracy. To address this situation, the combination of an inertial navigation system (INS) with GNSS has become an effective means to improve positioning and attitude determination accuracy and reliability.

[0003] An inertial navigation system (INS) can provide multi-dimensional navigation information such as position, velocity, and attitude, is not affected by external signal interference, and has good concealment and autonomy, making it an ideal independent navigation system. However, inertial navigation systems also have certain limitations. The error will gradually accumulate over time, which means that long-term use may lead to a decrease in accuracy. In addition, traditional inertial sensors are large in size, high in power consumption, and have certain applicability limitations for different types of carriers. However, with the continuous advancement of technology, micro-electromechanical system inertial measurement units (MEMS IMUs) have been able to provide high-precision measurements, and their cost has been significantly reduced, making high-precision inertial navigation systems gradually have a wider application prospect.

[0004] GNSS attitude measurement systems provide basic information about a vehicle, including its heading, track, attitude, position, and velocity. However, in traditional attitude measurement, one approach is to utilize the characteristics of the baseline vector to determine the vehicle's attitude parameters relative to a reference coordinate system. This method is primarily implemented through two approaches: one is to construct a cost function related to the baseline vector to estimate the optimal attitude solution, thus transforming it into the classic Wahba problem. However, this method faces certain difficulties in attitude accuracy analysis. Another approach is to use the baseline vector as a measurement value, obtaining the vehicle's attitude parameters through direct calculation or least-squares estimation. However, existing location-domain-based attitude measurement methods separate the calculation of the baseline vector and attitude parameters, neglecting their correlation, which may affect the reliability of attitude measurement to some extent. This is easily affected by signal interference and obstruction in complex environments, potentially amplifying the aforementioned problems and leading to decreased attitude accuracy or failure. Especially in urban areas with high-rise buildings, tunnels, or dense trees, the quality of GNSS signals is often severely challenged, affecting the stability and accuracy of attitude estimation.

[0005] Therefore, there is an urgent need for a method that can combine the characteristics of the urban environment and organically integrate the attitude angle information provided by the inertial measurement unit (IMU) with GNSS attitude measurement to improve the accuracy and robustness of attitude estimation. Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, the present invention aims to propose a COA attitude calculation method based on MEMS IMU-assisted search. This method utilizes the short-term accurate angle characteristics provided by MEMS IMU to calculate pitch and yaw angles to define the search boundary of COA, and provides high-precision attitude information through a search fitness function.

[0007] To achieve the above and other related objectives, this invention provides a COA attitude calculation method based on MEMS IMU-assisted search, which includes the following steps:

[0008] Step 1: Set up a MEMS IMU and BDS dual-antenna data acquisition platform, specifically as follows: Figure 3 As shown, in Figure 2 The data collected on the city streets shown is specifically from accelerometer, gyroscope, magnetometer, and BDS dual-antenna observations. The IMU used is JY901B, and the BDS dual-antenna data acquisition platform is built using TAU1312 and BT-800D. The equipment is magnetically attached to the top of the vehicle to collect dual-antenna data and inertial navigation data as the vehicle travels on the street.

[0009] Step two: Collecting inertial data using MEMS IMU to obtain pitch angle and yaw angle, which are used to determine the search boundary of COA pose optimization algorithm, specifically including:

[0010] 2-1 First, read the IMU raw data and preprocess it, converting angular velocity and acceleration to appropriate units. Although the car is driving on urban streets, the data quality remains stable due to the fact that the IMU itself is not affected by urban environmental factors such as tall buildings, tunnels, and tree shade. The main error sources come from the IMU itself. The errors of inertial measurement data mainly include three categories: noise (Bias and Noise), scale factor errors (Scale errors), and axis misalignments.

[0011] 2-2 To solve the error problem in inertial measurement data, the following methods are adopted: First, for noise and bias problems, Extended Kalman Filter (EKF) is used for dynamic error correction to improve data smoothness and accuracy. Second, for scale factor errors, the system calibration process is used to correct the scale factors of the IMU. Finally, through precise coordinate system alignment and IMU calibration, the deviation between axes is reduced.

[0012] 2-3 Through inertial data, the search boundary of the algorithm is determined. Let the yaw, pitch, and roll angles be z, y, and x, respectively. The calculation method using quaternions is as follows:

[0013]

[0014] 2-4 When using Extended Kalman Filter (EKF) to process IMU data, the system will make a prediction based on the current state at each time step and use the measurement data to update the prediction results. This method can effectively suppress the influence of noise, providing smoother and more accurate attitude estimation. Accurate attitude estimation helps to narrow the search space of the pose optimization algorithm and speed up the convergence speed.

[0015] Step three: Define the fitness function to be optimized based on the GNSS carrier phase observation model and the attitude observation model.

[0016] Further, step three specifically includes:

[0017] 3-1 The dual-antenna data collection model is shown in Figure 4 , where i and j represent the distribution of satellites in the sky, A and B represent the positions of the two receivers, is the baseline vector, p and h are the pitch angle and yaw angle of the baseline vector, and are the elevation angle and azimuth angle of the carrier plane of satellite j relative to the baseline.

[0018] 3-2 As described in 3-1, specifically, considering various noises during GNSS signal propagation, the carrier phase observation equation is modeled as follows:

[0019]

[0020] in, For carrier phase observations, Let λ be the carrier wavelength, r be the satellite-to-ground geometric distance, N be the integer ambiguity, I and T be the ionospheric and tropospheric errors, respectively, and dt and dT be the receiver clock error and satellite clock error, respectively. This is carrier phase observation noise.

[0021] 3-3 For the single-difference step of the carrier phase observations, since the baseline in the dual-antenna observation model is relatively short, the influence of tropospheric and ionospheric errors can be reasonably ignored, thereby simplifying the error model and improving solution efficiency and accuracy. The single-difference observation equation is:

[0022]

[0023] in, Here are the single-difference carrier phase observations, and f is the carrier frequency. For receiver clock bias, This refers to single-difference integer ambiguity. Differential analysis of the same satellite by two receivers at the same time can eliminate satellite clock bias.

[0024] 3-4 The single-difference data obtained in step 3-3 is further differentiald, that is, the two receivers perform differential analysis on the two satellites at the same time. Double-difference further eliminates receiver clock errors. The double-difference observation equation is:

[0025]

[0026] in, For s-difference carrier phase observations, and These are the line-of-sight vectors of the main antenna relative to the satellite. For the baseline vector, This represents double-difference integer ambiguity. By performing double-difference processing on the carrier phase observations, most errors can be eliminated, which benefits the subsequent solution process.

[0027] 3-5 Based on the dual-antenna observation model and double-difference observations, the satellite with the highest elevation angle is selected as the reference satellite, and the fitness function of the double-difference COA attitude optimization algorithm is constructed, specifically defined as follows:

[0028]

[0029] When h and p take the real values of heading angle and pitch angle respectively, the fitness function reaches its global maximum. At this time, the solution of the attitude angles is transformed into the optimal value search problem in the given interval. Due to the multi-peak characteristics of the fitness function of the attitude angles, there may be multiple local optimal solutions in the search interval, so it is necessary to effectively suppress the interference of local optimal solutions to the global optimal solution in the search process to ensure the accuracy and reliability of the attitude.

[0030] Step four: introduce an adaptive dynamic parameter adjustment mechanism to accelerate the convergence process of the COA algorithm and avoid falling into a local optimal solution.

[0031] Further, step four specifically includes:

[0032] 4-1 The basic idea of COA is to simulate two natural behaviors of long-nosed porcupines: (i) the behavior of hunting and attacking iguanas, and (ii) the behavior of escaping from predators. The implementation steps of COA are divided into two stages of exploration and development, and the detailed description of the exploration stage is as follows through a mathematical model.

[0033] In the initial stage of the implementation of COA, the position of the long-nosed porcupine in the search space is initialized by the following formula:

[0034]

[0035] F is the vector of the obtained target function value, is based on the target function value obtained by i long-nosed porcupines:

[0036]

[0037] The optimal value of the adaptive function is found by searching, and in the exploration stage, N long-nosed porcupines are divided into two groups, each with N / 2, and their positions are represented by the following two formulas and whether to update the position is judged.

[0038]

[0039]

[0040]

[0041] The new position of each long-nosed porcupine calculated will be accepted for updating if the target function value is improved, otherwise, the long-nosed porcupine will remain in the previous position. Wherein, j represents the decision variable corresponding to the p and h of the fitness function, represents the target value of the decision variable j, is the new position calculated by the i-th long-nosed porcupine; if the position is updated.

[0042] 4-2 The process of the second stage is mathematically based on modeling the natural behavior of raccoon dogs in escaping from the current position when encountering predators. In this strategy, the raccoon dog adjusts its position to approach the current state, and finally reaches a safe area, thus embodying the development capability of COA in local search.

[0043]

[0044]

[0045]

[0046] wherein, is the new position calculated for the ith raccoon dog in the second stage, and are the local upper bound and the local lower bound of the jth decision variable, respectively; if the position is updated .

[0047] 4-3 In order to improve the search accuracy of the algorithm in the later stage, the amplitude of random disturbance is gradually reduced during the running of the algorithm.

[0048]

[0049]

[0050] wherein, is the random disturbance amplitude of the tth iteration, is the initial random disturbance amplitude.

[0051] Step five: In order to solve the problem of too large jump or slow convergence speed in the solution space, a step size adaptive adjustment mechanism is proposed to increase the step size in the early stage of the algorithm to promote global search, and to reduce the step size in the later stage to improve the local development accuracy.

[0052] Further, step five specifically includes:

[0053] 5-1 The step size is dynamically adjusted in each iteration as follows:

[0054]

[0055]

[0056] wherein, is the step size of the tth iteration; is the step size factor, is the initial step size ratio, which can be adjusted according to needs.

[0057] 5-2 As described in 5-1, the step size is dynamically adjusted in each iteration, and is flexibly applied in different stages (such as the exploration stage) of the algorithm to optimize the performance of the search process:

[0058]

[0059] Step six: the fitness function is defined according to the carrier phase observation model and the attitude observation model in step three, the decision variable is p and h, the search range is obtained by solving the inertial data, and the improved COA is applied to search and optimize the attitude.

[0060] Further, step six specifically includes:

[0061] The fitness function to be processed is:

[0062]

[0063] The upper and lower bounds of the search are determined by the range of p and h solved on the basis of the inertial data, the convergence speed is accelerated, and thus the attitude is quickly and effectively solved.

[0064] Compared with the prior art, the advantages of the present application are:

[0065] The present application proposes a COA attitude solving method based on MEMS IMU assisted search, which fully utilizes the characteristics of MEMS IMU in providing high-precision attitude information in a short time, provides a reliable search space for the search process of the COA algorithm, thereby significantly improving the search efficiency, and realizing fast and efficient attitude solving. The method defines a fitness function for attitude solving search, and proposes a step size adaptive adjustment mechanism and a dynamic parameter adjustment strategy for the problem of large jump and slow convergence speed in the solution space. The method provides a fast and reliable attitude determination scheme for the navigation of the carrier in complex urban environment, and has broad application prospect and practical value. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to further illustrate the content described in the present application, the specific embodiments of the present application will be further described in detail below in combination with the drawings. It should be understood that these drawings are only typical examples, and should not be regarded as limiting the scope of the present application.

[0067] Figure 1 is a flow chart of the COA attitude solving method based on MEMS IMU assisted search.

[0068] Figure 2 is a scene diagram of collecting data in a city environment where GNSS signals are weak.

[0069] Figure 3This is a schematic diagram illustrating the specific implementation of data acquisition from MEMS IMU and BDS dual antennas.

[0070] Figure 4 It is a diagram showing the spatial geometry of the dual-antenna receiver and the satellite.

[0071] Figure 5 This is a search iteration graph of the COA attitude calculation method based on MEMS IMU-assisted search. Detailed Implementation

[0072] The following specific examples illustrate the embodiments of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. The specific embodiments of the present invention will be described below:

[0073] This invention provides a method for calculating COA attitude based on MEMS IMU-assisted search, and the specific embodiment includes the following steps:

[0074] The core idea of ​​this invention is to utilize MEMS IMU to acquire inertial navigation data, process the acquired data to obtain attitude information, and use this information as the upper and lower bounds for the COA algorithm search. Based on the spatial geometric relationship between the dual-antenna receiver and the satellite, and combining the GNSS carrier phase observation model and the attitude observation model, a fitness function to be optimized is constructed. To accelerate the convergence process of the COA algorithm and avoid getting trapped in local optima, an adaptive dynamic parameter adjustment mechanism is introduced. Furthermore, to address the problem of slow convergence speed due to large jumps in the solution space, an adaptive step size adjustment strategy is proposed. Within the spatial range defined by the inertial navigation system, the improved COA algorithm is used for attitude search and optimization, thereby significantly improving solution efficiency and attitude accuracy.

[0075] like Figure 1 The diagram shown is a flowchart of the COA attitude calculation method based on MEMS IMU-assisted search. The specific implementation process of this invention is as follows:

[0076] To address the aforementioned problems in the existing technology, the present invention aims to propose a COA attitude calculation method based on MEMS IMU-assisted search. This method utilizes the short-term accurate angle characteristics provided by MEMS IMU to calculate pitch and yaw angles to define the search boundary of COA, and provides high-precision attitude information through a search fitness function.

[0077] To achieve the above object and other related objects, the application discloses a COA attitude solution method based on MEMS IMU auxiliary search, and the method comprises the following steps.

[0078] Step one: build a MEMS IMU and BDS double-antenna data acquisition platform, as shown in the drawings, collect data in urban streets as shown in the drawings. Specifically, it contains acceleration, gyroscope, magnetometer and BDS double-antenna observation data, IMU selects JY901B, and the BDS double-antenna data acquisition platform is built through TAU1312 and BT-800D. The device is magnetically attracted to the top of the trolley, and the double-antenna data and inertial navigation data of the vehicle driving on the street are collected. Figure 3 Figure 2 Step two: use the MEMS IMU to collect inertial data to obtain the pitch angle and yaw angle, which are used to determine the search boundary of the COA attitude optimization algorithm, and the specific steps include:

[0079] 2-1 First, read the IMU raw data and perform preprocessing, and convert the angular velocity and acceleration into appropriate units. Although the trolley drives in urban streets, the data quality remains stable because the IMU itself is not affected by urban environmental factors such as high-rise buildings, tunnels, tree shadows and the like. The main error source comes from the IMU itself. The error of inertial measurement data mainly includes three categories: noise (Bias and Noise), scale factor error (Scale errors) and axis misalignment (Axismisalignments).

[0080] 2-2 To solve the error problem in the inertial measurement data, the following methods are adopted: first, for the noise and bias problem, the extended Kalman filter (EKF) is used for dynamic error correction to improve the smoothness and accuracy of the data. Secondly, for the scale factor error, the scale factors of the IMU are corrected through the system calibration process. Finally, through accurate coordinate system alignment and IMU calibration, the deviation between axes is reduced.

[0081] 2-3 Through the inertial data, the attitude solution is determined, and the algorithm search boundary is determined. Let the yaw, pitch and roll angles be z, y and x respectively, and the calculation mode of the quaternion is as follows:

[0082]

[0083]

[0084] ​​2-4 When using Extended Kalman Filter (EKF) to process IMU data, the system makes a prediction based on the current state at each time step and updates the prediction results using measurement data. This method effectively suppresses the influence of noise, thus providing a smoother and more accurate attitude estimation. Accurate attitude estimation helps to reduce the search space of the attitude optimization algorithm and accelerates the convergence speed.

[0085] Step 3: Define the fitness function to be optimized based on the GNSS carrier phase observation model and attitude observation model.

[0086] Furthermore, step three specifically includes:

[0087] 3-1 Dual-antenna data acquisition model as follows Figure 4 As shown, satellites i and j are distributed in the sky; A and B are the positions of two receivers. p is the baseline vector; p and h are the pitch and yaw angles of the baseline vector. and Let be the elevation angle and azimuth angle of satellite j relative to the carrier plane of the baseline.

[0088] 3-2 As described in 3-1, specifically, considering various noises during GNSS signal propagation, the carrier phase observation equation is modeled as follows:

[0089]

[0090] in, For carrier phase observations, Let λ be the carrier wavelength, r be the satellite-to-ground geometric distance, N be the integer ambiguity, I and T be the ionospheric and tropospheric errors, respectively, and dt and dT be the receiver clock error and satellite clock error, respectively. This is carrier phase observation noise.

[0091] 3-3 For the single-difference step of the carrier phase observations, since the baseline in the dual-antenna observation model is relatively short, the influence of tropospheric and ionospheric errors can be reasonably ignored, thereby simplifying the error model and improving solution efficiency and accuracy. The single-difference observation equation is:

[0092]

[0093] in, Here are the single-difference carrier phase observations, and f is the carrier frequency. For receiver clock bias, This refers to single-difference integer ambiguity. Differential analysis of the same satellite by two receivers at the same time can eliminate satellite clock bias.

[0094] 3-4 The single-difference data obtained in step 3-3 is further differenced, that is, the same time of two receivers is again differenced for two satellites, and through double difference, the receiver clock error can be further eliminated. The double difference observation equation is:

[0095]

[0096] wherein, is the s double-difference carrier phase observation value, and are the main antenna for satellite line-of-sight direction vectors, is the baseline vector, is the double-difference integer ambiguity. Through double-difference processing of the carrier phase observation value, most of the errors can be eliminated for subsequent solving process.

[0097] 3-5 According to the double-antenna observation model and the double-difference observation value, the maximum elevation satellite is selected as the reference satellite, and the double-difference COA attitude optimization algorithm fitness function is constructed, which is specifically defined as follows:

[0098]

[0099] When h and p take the true heading angle and pitch angle respectively, the fitness function reaches its global maximum value. At this time, the solution of the attitude angle is converted into an optimal value search problem in a given interval. Since the fitness function of the attitude angle presents a multi-peak characteristic, there may be multiple local optimal solutions in the search interval, so in the search process, the interference of local optimal solutions to the global optimal solution needs to be effectively suppressed to ensure the accuracy and reliability of the attitude.

[0100] Step four: introducing an adaptive dynamic parameter adjustment mechanism to accelerate the convergence process of the COA algorithm and avoid falling into a local optimal solution.

[0101] Further, step four specifically includes:

[0102] 4-1 The basic idea of COA is to simulate two natural behaviors of long-nosed raccoon: (i) the behavior of hunting and attacking iguanas, (ii) the behavior of escaping from predators. The implementation steps of COA are divided into two stages of exploration and development, and the exploration stage is described in detail through a mathematical model as follows.

[0103] In the initial stage of COA implementation, the position of long-nosed raccoon in the search space is initialized by the following formula:

[0104]

[0105] F is the vector of the obtained target function value, is the target function value obtained based on i long-nosed raccoons:

[0106]

[0107] The optimal value of the adaptive function is found by searching. In the exploration phase, N raccoon dogs are divided into two groups, each with N / 2, and their positions are represented by the following two equations and whether to update the position is determined.

[0108]

[0109]

[0110]

[0111] The new position of each raccoon dog calculated will be accepted for updating if the objective function value is improved, otherwise, the raccoon dog will remain in the previous position. Where j represents the decision variable corresponding to the p and h of the adaptive function, represents the target value of the decision variable j, is the new position of the ith raccoon dog calculated; if the position is updated.

[0112] 4-2 The process of the second phase is mathematically based on modeling the natural behavior of raccoon dogs fleeing their current position when encountering a predator. In this strategy, the raccoon dog adjusts its position to approach the current state, ultimately reaching a safe area, thus embodying the development capabilities of COA in local search.

[0113]

[0114]

[0115]

[0116] where, is the new position of the ith raccoon dog calculated in the second phase, and are the local upper and lower bounds of the jth decision variable, respectively. If the position is updated .

[0117] 4-3 To improve the search accuracy of the algorithm in the later stage, the amplitude of random disturbance is gradually reduced during the running of the algorithm.

[0118]

[0119]

[0120] where, is the random disturbance amplitude of the tth iteration, is the initial random disturbance amplitude.

[0121] Step five: In order to solve the problem of too large jump or slow convergence speed in the solution space, a step size adaptive adjustment mechanism is proposed, which increases the step size in the early stage of the algorithm to promote global search, and reduces the step size in the later stage to improve the local development precision.

[0122] Further, step five specifically includes:

[0123] 5-1 The step size is dynamically adjusted in each iteration as follows:

[0124]

[0125]

[0126] wherein, is the step size of the tth iteration; is the step size factor, is the initial step size ratio, which can be adjusted as needed.

[0127] 5-2 As described in 5-1, the step size is dynamically adjusted in each iteration, and is flexibly applied in different stages (such as the exploration stage) of the algorithm to optimize the performance of the search process:

[0128]

[0129] Step six: In step three, the fitness function is defined according to the carrier phase observation model and the attitude observation model, the decision variable is p and h, the search range is obtained by solving the inertial data, and the improved COA is applied to search and optimize the attitude.

[0130] Further, step six specifically includes:

[0131] The fitness function to be processed is:

[0132]

[0133] The upper and lower bounds of the search are determined by the ranges of p and h calculated on the basis of the inertial data, the convergence speed is accelerated, and the attitude is quickly and effectively solved.

[0134] Compared with the prior art, the advantages of the present application are:

[0135] This invention proposes a COA attitude determination method based on MEMS IMU-assisted search. It fully utilizes the high-precision attitude information provided by MEMS IMUs in a short time, offering a reliable search space for the COA algorithm's search process, thus significantly improving search efficiency and achieving fast and efficient attitude determination. The method defines a fitness function for attitude determination search and proposes an adaptive step size adjustment mechanism and a dynamic parameter adjustment strategy to address potential large jumps in the solution space and slow convergence speeds. This method provides a fast and reliable attitude determination scheme for vehicle navigation in complex urban environments, with broad application prospects and practical value.

Claims

1. A COA attitude calculation method based on MEMSMU-assisted search, characterized in that, The method includes the following steps: Step 1: Set up a MEMSIMU and BDS dual-antenna data acquisition platform to collect observation data including accelerometer, gyroscope, magnetometer, and BDS dual-antenna data; Step 2: Use the data collected by the IMU's accelerometer, gyroscope, and magnetometer to calculate the attitude, which serves as the boundary for the COA algorithm search; Step 3: Define the fitness function to be optimized based on the GNSS carrier phase observation model and attitude observation model; Step 4: Introduce an adaptive dynamic parameter adjustment mechanism to accelerate the convergence process of the COA algorithm and avoid getting trapped in local optima; Step 5: To address the problem of slow convergence due to excessive jumps in the solution space, an adaptive step size adjustment mechanism is proposed. Step 6: Within the attitude space defined by the inertial navigation system, apply the improved COA to perform attitude search and optimization, thereby improving the efficiency and accuracy of attitude calculation.

2. The COA attitude calculation method based on MEMSMU-assisted search according to claim 1, characterized in that, Step two is as follows: 2-1 Attitude calculation is performed using the AHRS attitude resolution algorithm, where yaw, pitch, and roll angles are represented by z, y, and x, respectively, and are expressed using quaternions as follows: 2-2 The IMU data is filtered by an extended Kalman filter (EKF). At each time step, the EKF predicts the system state and updates the state estimate and covariance matrix by combining the measurement data, thereby effectively reducing the impact of noise and providing smoother and more accurate attitude calculation results.

3. The COA attitude calculation method based on MEMSMU-assisted search according to claim 1, characterized in that, The specific steps are as follows: 3-1 Based on the GNSS carrier phase observation model and attitude observation model, the fitness function to be optimized, the carrier phase double-difference observation equation, is defined as follows: 3-2 Based on the dual-antenna observation model and double-difference observations, the satellite with the highest elevation angle is selected as the reference satellite, and the fitness function of the double-difference COA attitude optimization algorithm is constructed: Here, h and p represent the yaw angle and pitch angle, respectively, as decision variables.

4. The COA attitude calculation method based on MEMSMU-assisted search according to claim 1, characterized in that, Step four is as follows: An adaptive dynamic parameter adjustment mechanism is introduced to accelerate the convergence process of the COA algorithm and avoid getting trapped in local optima. The basic idea of ​​COA is to simulate two natural behaviors of the long-nosed raccoon: (i) hunting and attacking iguanas, and (ii) escaping predators. The implementation steps of COA are divided into two stages: exploration and development. In the second phase, the location update strategy is as follows: Wherein, is the new position calculated for the second stage of the i-th coati, and are the local upper bound and local lower bound of the j-th decision variable respectively; if F i P1 < F, then update the position To improve the search accuracy of the algorithm in the later stages, the amplitude of the random perturbation is gradually reduced during the algorithm's operation; 5. The COA attitude calculation method based on MEMSMU-assisted search according to claim 1, characterized in that, Step five is as follows: To address the issues of excessive jumps in the solution space or slow convergence, an adaptive step size adjustment mechanism is proposed, which increases the step size in the early stages of the algorithm to promote global search. The specific strategy is as follows: in, step t =β · (ub-lb), Step t β is the step size for the t-th iteration; t β0 is the step size factor, and β0 is the initial step size ratio, which can be adjusted as needed.

6. The COA attitude calculation method based on MEMSMU-assisted search according to claim 1, characterized in that, Step six is ​​as follows: In step three, a fitness function is constructed based on the carrier phase observation model and the attitude observation model, where the decision variables are p and h, and the search range is calculated from the inertial navigation data. Within the attitude space determined by the inertial navigation system, the improved COA algorithm is used for attitude search and optimization, which effectively accelerates the convergence process of the algorithm search and achieves fast and efficient attitude calculation while ensuring high accuracy.

Images