Title
Improved Attitude Determination Using Low-Cost MEMS Sensors
Attitude determination is an essential requirement in a wide range of applications like vehicle and space navigation, robotics, virtual environment, surveillance, and Unmanned Air Vehicle. In the standard navigation algorithms, attitude is computed by numeric integration of the gyroscopes (abbreviated as gyros). However, gyros specifically MEMS gyros suffer from large bias errors, which when integrated with time, introduce an unacceptable error in the computed attitude.
For static and low dynamic applications, attitude can also be computed by vector matching using two or more directional vectors. Three accelerometers and three magnetometers rigidly mounted in x, y and z axes can be used as directional vectors to sense gravity and Earth’s magnetic field vectors. As the reference vectors of
each of these are known, vector matching technique can be used for attitude computation using these sensors. However, the attitude computed via vector matching is noisy as the sensors noise directly appear in the computed attitude. Also, the attitude computed via vector matching is not reliable under accelerated motion and magnetic disturbance as in such cases, the accelerometers will sense linear acceleration in addition to the gravity and the magnetometers will measure nearby magnetic field in addition to the Earth’s field.
In the proposed research, we combined both methods and proposed techniques to compute improved attitude under static and dynamic motions. The first method proposed a simplified Quaternion feedback structure for gyro biases estimation and correction. This scheme uses gyros for attitude determination and a combination of accelerometers and magnetometers as aiding sensors for gyros bias errors estimation. The scheme functions in a closed loop by continuously estimating and correcting biases of the gyros. In the other two techniques, we used Kalman filter based data fusion in direct and indirect configurations. These techniques use gyros, accelerometers and magnetometers and fuse their data for error detection and correction. In indirect configuration, error states are estimated which are then used for correction. In the second method, attitude parameters are directly estimated in quaternion form using Kalman filter. Experimental tests have been performed to verify the proposed schemes under static and dynamic conditions using a self-developed setup consisting of MEMS-based IMU and FPGA based electronics.
Quaternion feedback configuration is also compared with Kalman filter based data fusion algorithm. The proposed quaternion feedback showed comparable results with low computational requirements and lesser probability of instability. It also caters for the effect of linear acceleration during which accelerometer outputs are not reliable for gravity vector sensing.
As in MEMS-based IMU, three-axis gyros, accelerometers and magnetometers are already available; the proposed methods can serve as self-aiding schemes for improved attitude determination.
<
>