A Rotor Unmanned Aerial Vehicle Load
with Attitude Estimation CombinationSystem
Qingtong Wu
School of Communication Engineering
HangzhouDianziUniversity
Hangzhou, China
e-mail:
Tianqi Liu, Xin Zheng
College of Electronic Information and Control Engineering
Beijing University of Technology
Beijing, China
e-mail:
Abstract—This paper designs a combination solution reference system which is applicable to the rotor unmanned aerial vehicle (UAV) with attitude load.The system could fuse Micro-Electro-Mechanical System(MEMS), dual-antenna Global Positioning System(GPS) and geomagnetic information through a complementary filter.Ensuring the accuracy of gyro and accelerometer output data through the precise calibration of MEMS inertial devices,combining the attitude output of dual antenna GPS and geomagnetic, the paper designs a integrated attitude calculationhardware and software system and realizes high precision calculation of the Rotor UAV composite attitudesystem.
Keywords—UAV，attitudeestimation，complementary filter
I.Introduction
Rotor UAV has a broad prospect in military and civilian due to its outstanding flight performance[1]. And the study of its load attitude has become a hot issue today. Attitude calculation based on the MEMS inertial devices has been studied and used widely. But because of the gyroscope have temperature drift, and the constant error accumulated over time, the acceleration sensor have great influenced by the engine block vibration in dynamic flying process[2]. For the dual antenna GPS satellite navigation position and orientation system, all-weather continuous positioning and orientation could be completed in the outdoor, and the accuracy is not follow the change of time, but in a sheltered environment, GPS was limited because it cannot obtain valid data[3]. The geomagnetic sensor has the characteristics of refers to the North precisely, but susceptible to electromagnetic interference. Therefore, combination of MEMS inertial devices/ dual antenna GPS/ geomagnetic sensor, combines the advantages of the specific environments, to achieve the composite attitude calculation of low cost, all-weather, multi environment, anti-interference, high precision. By establishing the mathematical model of MEMS inertial devices and the calibration of installation error, to obtain the accurate output information from gyroscope and accelerometer.Completed multi sensor attitude acquisition, design the airborne MEMS/ dual antenna GPS/ geomagnetic composite attitude solver based on complementary filter, which can output precision attitude data long-term and stability. Comparison of the IMU sensor in YS-X6 flight control system, which is produced by ZERO-TECH Company, could satisfy the rotor UAV flight requirements and low cost application requirements.
II.system composition structure
The system is composed by 3-axis MEMS inertial devices, dual antenna GPS, geomagnetic sensor, digital signal processor (DSP) and other peripheral circuit. In which the main system with the Blackfin518 DSP processor of ADI company, to complete the data acquisition, fusion and filtering, its MEMS inertial unit is a piece of 3-axis accelerometer and high precision sensor SCC1300 with single axis gyroscope, which is produced by Murata Electronics Oy (The original company name is VTI Technologies), three pieces constitute the 3-axis gyroscope structure that can output data through a serial peripheral interface (SPI). Meanwhile, the system board equipped with YA529 geomagnetic sensor made by Yamaha Corporation and output data through the integrated circuit bus interface (IIC). Besides, it connected with the dual antenna GPS on slave system board through serial port, the slave system board uses ADU3601 module made by Beijing StarNeto Technology Co, the orientation precision is 0.15 degrees when baseline length is 1 meter. This system of combination structure has small and convenience characteristics, the block diagram shown in Figure 1.
Fig. 1. Block diagram of hardware system
III.MEMS inertial devices Calibration
A. Coordinate instructions
(1)Chip coordinates
The coordinate does not meet the right-hand rule, as shown in Figure 2.
Fig.2. Chip coordinates
(2)MEMS INS Coordinate
The coordinates is established by three sensor system shown in Figure 3.
Fig. 3. MEMS INS Coordinate
B. Gyroscope Calibration
(1)Mathematical model
Through access to large amounts of data and computing test, although two order gyroscope calculation model is more complex calculation than first-order equation, but the precision can be improved, so as to achieve the requirement of accuracy[4]. MEMS gyroscope output mathematical model in strap-down inertial navigation is:
(1)
In the formula, Wx, Wy, Wz are the outputs of MEMS gyroscope, wx0, wy0, wz0 are the drift values of MEMS gyroscope, wx, wy, wz are the inputs of MEMS gyroscope, Skx, Sky, Skz are scale factor of MEMS gyroscope,Kxy, Kxz, Kyx, Kyz, Kzx, Kzy, Kx, Ky, Kz are installation error coefficients of MEMS gyroscope.
By formula (1) can be derived that the equation of scaling factor, drift value and installation error coefficients is:
(2)
Put the value into w and W, the installation error coefficients Kx, Ky, Kz, the scale factor Skx, Sky, Skz, and the drift value wx0, wy0, wz0 can be obtained by polynomial fitting.
By the X-axis MEMS gyroscope calibration, we get Y-axis and Z-axis output from formula (1).
(3)
Similarly, we can derive the installation error coefficients Kyx, Kzx and drift value wy0, wz0 by polynomial fitting. In the same way, by the Y-axis MEMS gyroscope calibration could obtain installation error coefficients Kxy, Kzy and drift value wx0, wz0, and by the Z-axis MEMS gyroscope calibration could obtain installation error coefficients Kxz, Kyz and drift value wx0, wy0.
However, the experimental results shows that the obtained parameters by polynomial fitting for (2) and (3) cannot achieve the optimal effect, therefore this paper designs a more elaborate method. According to the experimental data of theplus or minus angle, it can be derived to formula (4):
(4)
Add the two equations in formula (4) together to get formula (5):
(5)
At this time we get Kx and wx0 by polynomial fitting, similarly get Ky, wy0 and Kz, wz0. As can be seen in the formula (5) which to eliminate the influence of Sk which brings to the fitting, and average error of plus or minus angle.
Finally, we derived the average of wx0、wy0、wz0 by formula (4) and (5), so that the drift value can be compensated to obtain more accurate bias. The above mentioned process accomplishes the calculation of gyro calibration parameters, and improves the precision of calculated angle.
By solving the equations we obtained the calibrated angle, the data error is 0.06 DEG /s through statistic. For example to around the Y axis to -60 DEG/s rotations, the MATLAB fitting chart shown in Figure 4, figure 5, Figure 6.
Fig. 4. Data fitting of X gyroscope
Fig. 5.Data fitting of Y gyroscope
Fig. 6.Data fitting of Z gyroscope
(6)
In this formula, Ax,Ay,Az are the actual measured values of MEMS accelerometer which unit for g, ax0,ay0,az0 are the bias of MEMS accelerometer which unit for g, Ka1,Ka2,Ka3 are the installation error coefficients of MEMS accelerometer, Kax3,Kay3,Kaz3 are the error coefficients relative to quadratic equation of MEMS accelerometer which unit for g-1, Sax,Say,Saz are scaling factor of MEMS accelerometer.
(2)24 position calibration of MEMS accelerometer
The calibration is achieved in the horizontal rotary table, this paper improved the 6 position method and put forward a kind of 24 position calibration of MEMS accelerometer in order to counteract the horizontal error.This method would deal 6 position tests in all directions respectively through rotary table, then according to the experimental method and the order of 24 position directions, to divide 4 groups of directions X(g),X(-g),Y(g) ,Y(-g),Z(g),Z(-g).
The output of X-axis accelerometer shown in formula (7), and could derive the output of Y-axis accelerometer and Z-axis accelerometer Ay1,Ay2,Ay3,Ay4,Ay5,Ay6;Az1,Az2,Az3,Az4、Az5,Az6 in same way.
(7)
The coefficients of X-axis MEMS accelerometer module shown in formula (8), and could derive the coefficients of Y-axis and Z-axis MEMS accelerometer module ay0,Say,Kay1,Kay2,Kay3; az0,Saz,Kaz1,Kaz2,Kaz3 in the same way.
(8)
According to above process we can get six positions X(g),X(-g),Y(g),Y(-g),Z(g),Z(-g), four groups in total, calculate them separately and take the average value, and get a group of values, which includes bias, scaling factor and installation error coefficients ,that is more precise than single value.
By solving the equation group,we obtain the calibrated-acceleration of the system, and the data error is 0.0009/g. Figure7,8,9 show an example of MATLAB fitting curve of all data pointing to the earth in the direction of Y axis.
Fig.7. Data fitting of accelerometer X
Fig. 8. Data fitting of accelerometer Y
Fig. 9. Data fitting of accelerometer Z
IV.Attitude calculation of multi sensor combination
There are a lot of methods of multi-sensor data fusion, such as neural network, wavelet analysis, Kalman filtering attitude solution algorithm. This paper select the complementary filter for processing, it has lower computation complexity, lower requirements for the processor performance, and has been widely used in combined navigation system.
A. Data fusion method of composite attitude calculation system
Through the MEMS gyroscope attitude solution and accelerometer, dual antenna GPS, data output of geomagnetic sensor, using complementary filter attitudeestimation method, to complete the output of attitude calculation system, to get more accurate position information. The schematic diagram of composite attitude calculation system is shown in Figure 10.
Fig. 10. composite attitude calculation system
B. Strapdown matrix
While calculating the attitude, it will be required to convert the data of vector coordinate system into navigation coordinate system,according to the quaternions method for solving the Runge-Kutta equation, to get attitudematrix , which is the relationship between T and three attitude angles, in which n stand for navigation for North East ground coordinate system[6].
(9)
According to the relationship between the platform and the vector coordinate system, is the function of, set, then:
(10)
C. The basic principle of complementary filter
Complementary filter distinguish the noise characteristics through frequency domain, its principle is to filter out the high frequency part from the signal containing the high-frequency noise through a low-pass filter, and then filter out the low frequency part from the signal containing the low-frequency noise through a high-pass filter, to get a better and complete signal[7].
Set the real attitude angle is R, R0 is aattitude angle with high-frequency noise u, then R0=R+u, R1 is aattitude angle with low-frequency noise v, then R1=R+v. The first-order high pass filter F0(s)=s/(s+K), the first-order low pass filter F1(s)=K/(s+K). It’s known as F1(s)+F2(s)=1.
(11)
The schematic diagram is shown in Figure 11.
Fig. 11. schematic diagram of complementary filter
D. The attitude calculation based on complementary filter and quaternions
Complementary filter usually use proportional integral (PI) compensation method. Use, the proportional gain designs the cut-off frequency of filter, the integral gain designs the time of eliminate static error. The schematic diagram of complementary filter attitude calculation system is shown in Figure 12.
Fig. 12. schematic diagram
is the optimal estimated value of complementary filter for gravity acceleration, is the gravity acceleration of the accelerometer testing result, the cross product is the error of pitch and roll angle , is the geomagnetic measurement magnetic field vector, is the complementary filter estimated magnetic field vector, the cross product is the error of yaw angle , is the baseline vector of dual antenna GPS, is the best estimated value of GPS baseline vector to complementary filter.
(12)
(13)
(14)
In which is the compensation dosage of gyroscope drift value that generated by error e through a proportional integral module, is the gyroscope measured angular velocity, is the corrected angular velocity by fusion accelerometer, geomagnetic and dual antenna GPS data. Introducing into the Runge-Kutta to update the quaternions equation can be obtained exactly quaternions with time passing, then get the attitude matrix and attitude angle.
V.experimental results
In order to verify the performance of the combination attitude calculation system based on complementary filter, it is equipped with UAV which is made by ZERO-TECH Company for experiment, to fixed in the same horizontal with ZERO-TECH YS-X6 flight control system, its flight control system using high performance processor and integrated sensor, the attitude data static precision is ±0.2º, the dynamic precision is ±2º. By operating the UAV to complete the in-situ attitude transformation motion, and to collect the both data by computer serial port, then use MATLAB to complete the drawing of the movement trend as shown in Figure13,14,15.
Experimental results show that, when the UAV at large rotation speed, the largest tracking error of pitch angle is ±0.9°, the largest tracking error of roll angle is ±1.5°, the largest tracking error of yaw angle is ±1.0°. In the vast majority of time, the attitude angle tracking error of this system is less than ±1.5°. Compare to YS-X6 system, this system has a better dynamic precision.
Fig. 13. Pitch angle measurements
Fig.14. Roll angle measurements
Fig. 15. Yaw angle measurements
VI.physical system FIGURES
The physical system figures are as follows.
Fig. 16. Main board hardware
Fig. 17. GPS hardware
Fig. 18. system hardware
VII.conclusions
This paper studies on the rotor UAV payload attitude solution reference system, designs a kind of composite attitude calculation system which is based on complementary filter MEMS/dual antenna GPS/geomagnetic sensor. To achieve the precise calibration of MEMS inertial devices, and experimental results show that this system could fuse the multi-sensor data effectively and output stable attitude angle,it can satisfy the general payload demand of rotor UAV.
References
[1] Emanuel Stingu and Frank L. Lewis. A Hardware Platform for Research in Helicopter UAV Control[J]. Unmanned Aircraft Systems, 2009.
[2]Ren W.Zhang H Y.et al. A research on calibration of low-precision MEMS inertial sensors[C]//Proc. of the Control and Decision Conference,2013:3243-3247
[3]Zhu Y H,Cai T J, Yang Z P, Implementation of MEMS-IMU/GPS integrated navigation system[J]. Journal of Chinese Inertial Technol-ogy,2009,17(5);552-561.
[4]Lijun Song, Yongyuan Qin. A practical calibration method for MEMS gyroscope. Piezoelectric and acousto-optic, 2010.6
[5]Lijun Song, Yongyuan Qin. MEMS accelerometer six position test method. Measurement and control technology. 2009(7)：11—13
[6]Zhe Chen. The principle of strapdown inertial navigation system. Beijing: Aerospace press. 1986
[7] Baerveldt A J,Klang R.A low-cost and low-weight attitude estimation system for an autonomous helicopter [C]//Proc of Int ‘ IConf on Intelligent Engineering System,1997:391-395