|
|
1.INTRODUCTIONInertial measurement systems are widely used due to their wide application range, good dynamics, and high accuracy of data acquisition. For example, in small aircraft, they are used to obtain aircraft attitude information. With the development trend of society, small unmanned aerial vehicles (UAVs) are playing a huge role in surveying and mapping, three-dimensional real scene, land planning, public security maintenance, disaster monitoring, forestry investigation, smart water conservancy, digital city, emergency rescue and other fields. [1][2] Due to the wide application of UAVs, it is inevitable that they will be used in the harsh environment with extremely poor signals and low visibility, such as high altitude, fog, heavy rain, heavy snow, etc., which leads to the low operation efficiency of UAVs. Therefore, higher requirements are put forward for the attitude information of the UAV. Usually, a filter algorithm is used to solve the attitude information to obtain high-precision, high-reliability and high-stability attitude information. Julier et al. (2000) [3] used a filter better than extended Kalman filter for filtering, and the accuracy effect is easier to achieve than EKF or Gaussian second-order filter. However, this method requires a lot of computation and is not easy to implement. Shin et al. (2004) [4] developed a low-cost unscented Kalman filter, a quaternion-based UKF method, which allows a large initial attitude error. For the inertial navigation system with a fast increase in heading error, the processing result is better, but the calculation is large and EKF is better than UKF when the initial attitude error is small. Sabatini et al. (2006) [5] designed and tested a quaternion-based orientation filter, established a specific state vector model, and introduced the measurement noise covariance matrix to prevent the influence of magnetic field disturbance on the measurement accuracy, thus improving the accuracy of positioning estimation. Rong Zhu et al. (2007) [6] proposed a novel estimation method based on the Kalman fusion algorithm, which adopts a new state vector containing three orthogonal components of magnetic field and gravity, replacing the traditional Quaternions and Euler angles, thus establishing a simple linear Kalman model, the method is greatly simplified in computational complexity, and can be executed at a faster speed using an inexpensive microprocessor, which verifies the effectiveness of the method. Silson et al. (2011) [7] proposed a new fast initial coarse alignment estimation method for ship strapdown inertial attitude reference system. The advantage of this method is that it can deal with attitude errors effectively. The disadvantage is that only inertial measurement units are used for alignment, and sensor errors are not considered. Zou et al. (2014) [8] design of the improved Extended Kalman Filtering algorithm, using the fourth order runge kutta method to correction of gyro Angle, so as to correct the error of accelerometer and magnetometer data, estimate the process noise covariance of Kalman Filtering, so as to accurately locate gesture, but ignored the interference of the acceleration of gravity. Julier et al. (2014) [9] surveyed the state of the art in unscented techniques for nonlinear estimation and provided a number of examples to verify its advantages over the traditional linearization method, but there is a lack of certain experiments and evidence. Sheng et al. (2015) [10] proposed an algorithm for fusion of Multi sensor data, which was implemented by the strapdown AHRS hardware platform. The attitude angle can be corrected by PID control parameters and the attitude estimation is processed by the direction cosine matrix model. Zhao et al. (2015) [11] proposed the CKF algorithm, which uses the noise statistical estimator and has high adaptive robustness, this algorithm solved the problems of accuracy degradation and divergence under the condition of model uncertainty or prior noise statistics, and improved the robustness and adaptability of SCKF algorithm. Wang et al. (2017) [12] proposed a volumetric Kalman filter algorithm and deduced an augmented volumetric Kalman filter to deal with the strong nonlinearity of the INS model, effectively reducing the non-additive value in inertial measurement. The method is better than the traditional EKF/UKF method. Lu et al. (2020) [13] proposed an aircraft attitude information fusion method based on Extended Kalman Filter (EKF) to eliminate the influence of magnetic interference on attitude and heading estimation, but in the process of attitude determination, the linearization step of the system equation uses the Jacobian matrix to update the state transition matrix and the observation matrix, which has poor stability and large estimation deviation. Liu et al. (2022) [14] proposed an Adaptive Unscented Kalman Filter (AUKF), which uses the gradient descent method to optimize the key parameters of the unscented Kalman filter and without calculating the Jacobian matrix. The convergence speed is strengthened, and the attitude calculation accuracy and stability are very good, but the calculation amount is large, and the estimation accuracy is also greatly affected in complex working environments. G. Zhou [15-20] has also done a lot of research in this field. Considering the accuracy of attitude calculation, the influence of sensor errors and the algorithm requirements, a SOA-optimized Adaptive Extended Kalman Filter algorithm is proposed. This method optimizes the parameters R and Q of the filter, and finds the exact values of the two parameters, so as to achieve the purpose of solving. The algorithm has the advantages of the fast global convergence, the fast operation speed and the simple implementation. Through the static and the dynamic experiments, the advantages of the algorithm and the traditional Kalman filter algorithm are compared and analyzed. The algorithm improves the accuracy and performance of attitude solution. 2.OVERALL IMPLEMENTATION METHODThe SOA genetic algorithm is used to adaptively adjust the noise model parameters of the extended Kalman filter system to meet the requirements of the EKF algorithm model for accurate modeling. Since the model parameters of this filter are constantly changing over time, the optimization period of the genetic algorithm can be appropriately selected to reduce the calculation amount to a certain extent and optimize the filtering accuracy; through optimizing the function of SOA genetic algorithm, the optimal large-scale variation of parameters can be found, so as to achieve high-precision attitude calculation. The SOA genetic algorithm can iteratively obtain accurate parameters, which are the two key values R and Q of the extended Kalman filter, that is, the measurement noise value and the system noise value. Of course, all of these require repeated debugging and research through programming to get better results. The idea flow of the whole attitude calculation is shown in the following Fig. 1. 3.EXPERIMENT AND RESULT ANALYSISThis experiment is based on the Seagull Genetic algorithm [21] using the Extended Kalman Filter to solve the adaptive filtering. In the Windows 10 environment, the experiment is carried out through MATLAB R2014b, and the experimental situation is as follows. 3.1Experimental PlatformIn order to illustrate that the method proposed in this paper has certain advantages, the accuracy of attitude calculation is improved to a certain extent. In this paper, a flight experiment platform based on the INS-DH-OEM inertial navigation system is built, and the UAV dynamic flight experiment is carried out outdoors. This experiment uses the INS-DH-OEM inertial navigation system, which uses an advanced dual-antenna GNSS receiver, Honeywell HG4930 IMU, advanced MEMS accelerometer with 3-axis calibration full operating temperature range and a new generation of tactical grade MEMS gyroscope. The parameters of the inertial measurement unit are as follows: the serial port baud rate is 115200bps, the data output rate is 1~200HZ, the pitch/roll accuracy (static) is: 0.05°RMS, and the pitch/roll accuracy (dynamic) is: 0.5° RMS. High-precision inertial measurement unit (including gyroscope and accelerometer), as shown in the following Fig. 2. 3.2The ExperimentThe data of this experiment is from June 25, 2022, in front of the library of Yanshan Campus of Guilin University of Technology, Guilin City, Guangxi Zhuang Autonomous Region. using an octa-rotor UAV equipped with INS-DH-OEM sensor to carry out the flight experiment, data collection is read by the INS-DH-OEM high-precision inertial navigation system, and the data read by the sensor system within a certain period of time is selected for comparison with the attitude data obtained by the experimental method. 3.3Static ExperimentThe eight-rotor UAV is placed on the level ground, and the ideal value of the pitch angle and roll angle is 0°, as shown in the following Fig. 3. This experiment outputs normalized acceleration data, and the angle data is converted into radians and angles. After the Adaptive Extended Kalman Filter processing, the curve change graph of the attitude angle within 180 seconds is output. The actual attitude curve of the three axes is shown in the following Fig. 4. Through the Adaptive Extended Kalman Filter processing of the attitude data collected in the static state, it is found that compared with the ideal value of the roll angle of 0°, the actual roll angle deviation is 0.09°, and the algorithm has higher Compared with the ideal value of the pitch angle of 0°, the actual pitch angle deviation is 0.045°, and the calculation accuracy of the algorithm is high; since the magnetometer data is not included in the POS, the attitude calculation of the heading angle depends on Accelerometer and gyroscope, but between 20 seconds and 40 seconds, the error of accelerometer and gyroscope fluctuates greatly, so the calculated heading angle fluctuates greatly around 20 seconds, and basically remains stable in other time periods. 3.4Dynamic ExperimentLet the UAV equipped with the POS experimental platform change its attitude, and select the attitude measurement data under dynamic conditions. Experiment: collect the dynamic POS data within 100 seconds, use the Adaptive Extended Kalman Filter algorithm to calculate the attitude, and compare the three-axis attitude angle calculated within 100 seconds with the error of the real three-axis attitude angle obtained by the UAV with the SOA platform installed, as shown in Fig. 5. Analysis: Comparing the calculation result of the Adaptive Extended Kalman Filter algorithm with the real attitude angle, the deviation between the filter angle of roll and the real angle is within 0.2°; the deviation between the filter angle of pitch and the real angle is within 0.5°; and because of the POS does not contain magnetometer data, so the magnetic field data cannot be obtained. The true heading angle of the magnetic field is obtained by the fusion of the three-axis accelerometer and the gyroscope. The yaw filtering angle calculated by the Adaptive Extended Kalman Filter algorithm is not representative and cannot be compared. Therefore, in general, the real attitude angle curve changes can be accurately tracked by the Adaptive Extended Kalman Filter algorithm, but the value of the heading angle is very drifting and is not representative. 4.CONCLUSIONAiming at the attitude solution problem of octopteral drone, in order to solve the low cost and high precision requirements of attitude sensor installed on UAV, in this paper, an attitude solving algorithm based on SOA algorithm and Adaptive Extended Kalman Filter is proposed. The method uses the SOA algorithm to adjust the parameters R and Q of the Extended Kalman Filter, optimizes the state prediction and estimation, and effectively reduces the sensor drift and noise errors in the attitude solution process. Through the static data and dynamic data collected by the UAV platform equipped with imu, the attitude accuracy of the proposed method is verified, and the experimental results are analyzed and compared in detail. Experimental results show that, compared with the traditional algorithms, the algorithm improves the long-term stability and accuracy of attitude solution. REFERENCESMaria de Fatima Bento,
“Unmanned Aerial Vehicles: An Overview,”
Inside GNSS, 1 54
–61
(2008). Google Scholar
Xu sheng Lei, Jingjing Li,
“An adaptive navigation method for small unmanned aerial rotorcraft under complex environment,”
Measurement, 46
(10), 4166
–4171
(2013). https://doi.org/10.1016/j.measurement.2013.06.040 Google Scholar
Julier, S., Uhlmann, J. Hugh,
“F. A New Method for the Nonlinear Transformation of Means and Covariances in Filters and Estimators,”
IEEE Transaction on Automatic Control, 3 477
–482
(2000). https://doi.org/10.1109/9.847726 Google Scholar
Shin, E. and El-Sheimy, N.,
“An Unscented Kalman Filter for In-Motion Alignment of Low-Cost IMUs,”
in Position Location and Navigation Symposium,
273
–279
(2004). Google Scholar
A. M. Sabatini,
“Quaternion-based extended Kalman filter for determining orientation by inertial and magnetic sensing,”
in IEEE Transactions on Biomedical Engineering, 53
(7), 1346
–1356
(2006). https://doi.org/10.1109/TBME.2006.875664 Google Scholar
Rong Zhu; Dong Sun; Zhaoying Zhou; Dingqu Wang,
“A linear fusion algorithm for attitude determination using low-cost,”
MEMS-based sensors, 40
(3), 322
–328
(2007). Google Scholar
Silson, P.M.G,
“Coarse alignment of a ship’s strapdown inertial attitude reference system using velocity loci,”
IEEE Transactions on Instrument and Measurement, 60
(6), 1930
–1941
(2011). https://doi.org/10.1109/TIM.2011.2113131 Google Scholar
Zou Bo, Zhang Hua, Jiang Jun,
“Improved Extended Kalman Filter Attitude Determination Based on Multi-sensor Information Fusion [J],”
Computer Application Research, 31
(04), 1035
–1038+1042
(2014). Google Scholar
S. J. Julier and J. K. Uhlman,
“Unscented filtering and nonlinear estimation,”
in Proceedings of the IEEE,
401
–422
(2004). Google Scholar
H. Sheng, T. Zhang,
“MEMS-based low-cost strap-down AHRS research,”
Measurement, 59 63
–72
(2015). https://doi.org/10.1016/j.measurement.2014.09.041 Google Scholar
Zhao Liqiang, Wang Jianlin, Yu Tao, Jian Huan, Liu Tangjiang,
“Design of adaptive robust square-root cubature Kalman filter with noise statistic estimator[J],”
Applied mathematics and computation, 256352
–367
(2015). Google Scholar
Dingjie Wang, Lv H, Jie W.,
“Augmented Cubature Kalman filter for nonlinear RTK/MIMU integrated navigation with non-additive noise[J],”
Measurement, 97111
–125
(2017). Google Scholar
Lu Yanjun, Chen Yudi, Zhang Xiaodong, Zhang Taining,
“Research on attitude information fusion method based on extended Kalman filter [J],”
Journal of Instrument and Instrument, 41
(09), 281
–288
(2020). Google Scholar
Liu Kangan, Zhang Weiwei, Xiao Yongchao, Ye Mu,
“Attitude solution of quadrotor UAV based on adaptive unscented Kalman filter [J/OL],”
Electro-Optics and Control, 1
–7
(2022). Google Scholar
G. Zhou and R. Li,
“Accuracy evaluation of ground points from high-resolution satellite imagery IKONOS,”
Photogrammetry Engineering & Remote Sensing, 66
(9), 1103
–1112
(2000). Google Scholar
G. Zhou, C. Song, and W. Schickler,
“Urban 3D GIS from LIDAR and digital aerial images,”
Computers and Geosciences, 30
(4), 345
–353
(2004). https://doi.org/10.1016/j.cageo.2003.08.012 Google Scholar
G. Zhou, W. Chen, and J. Kelmelis,
“A comprehensive study on urban true orthorectification,”
IEEE Trans. on Geosciene and Remote Sensing, 43
(9), 2138
–2147
(2005). https://doi.org/10.1109/TGRS.2005.848417 Google Scholar
G. Zhou,
“Near Real-Time Orthorectification and Mosaic of Small UAV Video Flow for Time-Critical Event Response,”
IEEE Trans. on Geoscience and Remote Sensing, 47
(3), 739
–747
(2009). https://doi.org/10.1109/TGRS.2008.2006505 Google Scholar
G. Zhou and Vine Ambrosia, Albin Gasiewski, and G. Bland,
“Foreword to the Special Issue on Unmanned Airborne Vehicle (UAV) Sensing Systems for Earth Observations,”
IEEE Trans. on Geoscience and Remote Sensing, 47
(3), 687
–689
(2009). https://doi.org/10.1109/TGRS.2009.2013059 Google Scholar
G. Zhou, and L. Wang,
“Integration of GIS and Data Mining Technology to Enhance the Pavement Management Decision Making,”
J. of Transportation Engineering, 136
(4), 332
–341
(2010). https://doi.org/10.1061/(ASCE)TE.1943-5436.0000092 Google Scholar
Han Yi, Xu Zibin, Zhang Liang, Deng Lili,
“A New Foreign Intelligent Optimization Algorithm—Seagull Optimization Algorithm [J],”
Modern Marketing (Management Edition),
(10), 70
–71
(2019). Google Scholar
|
