Development of High-precision Navigation Algorithm by Fusion of GPS and MEMS Sensors

Y. Kosaka, T. Tsuchiya, M. Naruoka, H. Tomita, H. Kurihara, K. Ishida, T. Ichikawa

Abstract:	This paper presents a development of highly accurate navigation algorithm by sensor fusion, integrating information from civil-use GPS receiver and several MEMS sensors. Recently, more and more unmanned systems, such as robots and UAV (Unmanned Aerial Vehicle), have been developed as tools that support cultural life of human beings. It is necessary to obtain basic information of objects, such as position, velocity and attitude, for controlling and observing the moving objects. By applying navigation technology sophisticated in the field of aeronautics and aerospace, it is possible to get the state information. However, traditional navigation devices are intended to be used in the field where accuracy is the most important and they are too big, heavy and expensive to be generally used towards small moving objects. It is necessary to be small, light and low-cost for a navigation equipment to be used generally. This study is intended to develop an algorithm for a navigation device used in a moving object whose size is less than one meter cubic. In order to satisfy this demand, the INS/GPS integrated navigation algorithm has been studied in previous researches. This algorithm makes it possible to fuse information obtained with small, light and low accurate MEMS sensors and to develop small integrated navigation equipment. In our previous research, a generic integrated navigation device named “Sylphide” was developed by Naruoka et al. This device consists of MEMS inertial sensors (gyroscope, accelerometer) and a civil-use GPS receiver. It overcame problems of low accuracy and error accumulation resulting from stand-alone use of MEMS sensor. However, the previous researches are insufficient in terms of robustness, reliability and accuracy. It is especially problematic that estimation error of severe conditions, such as when GPS signals are not available or when a motion of the object is small, is not considered. Therefore, we propose a new algorithm, which enhances the existing INS/GPS integrated navigation algorithm by integrating more kinds of sensors. Two kinds of three-axis gyroscopes, two-kinds of three-axis accelerometers, three-axis geomagnetic sensor, barometer and GPS are utilized. Here, the word of “two kinds of sensors” means that two sensors detect the same quantity, where one sensor has higher resolution but less range width, and the other has lower resolution but larger range width. The key innovative step of this study is these two different gyroscopes and accelerometers are respectively used for detecting the same quantity of angular velocity and acceleration. It results in quality improvement of the estimations. Combining one measurement with the other by using high or low pass filters, measured values overall are more accurate and, as a result, the reliability of the whole system will be enhanced even though in the severe conditions. In addition, the geomagnetic sensor and barometer improve estimation accuracy of attitude and vertical information, respectively, compared with the original INS/GPS algorithm. In order to integrate sensor information, the EKF (Extended Kalman Filter) is applied. State values consist of velocity, position on the earth, altitude, attitude and bias of accelerometers, gyro scopes and a geomagnetic sensor. Here, altitude and attitude are described in quaternion representations. Inputs consist of acceleration obtained from an accelerometer, angle velocity obtained from a gyroscope, gravity obtained from the gravity model of WGS84 (the World Geodetic System 1984) and white noise affecting each sensor. Observed values consist of velocity, position and altitude obtained from a GPS receiver, and geomagnetic vector obtained from a geomagnetic sensor. The composition of these values may change due to future sensor integration. The EKF is composed of time update, where the inputs are obtained and the state values are processed according to the equation of motion, and measurement update, where the observed values are gathered and errors are estimated to perform correction. The functionality of the algorithm is preliminary validated with computational simulations with aircraft navigation data actually obtained in flight experiments. The data is collected with an ultra-precise navigation device, MSAS-GAIA. MSAS-GAIA is developed by JAXA (Japan Aerospace Exploration Agency) and it has already been proven to be precise enough to be utilized as a reference, for example, its attitude angle error is less than 0.1 degrees. In the simulations, the data measured by MSAS-GAIA are used as true state values, and the values made by adding arbitrary noises and biases to them are used as observed values of the sensors. Comparing the true state values with the values estimated by using the pseudo sensor information, the proposed algorithm is validated. Even though in severe condition, where there are a lot of noise and bias, estimation errors of all state values decrease more than previous algorithms. There are other interesting results obtained by integrating a geomagnetic sensor. Here, two cases are compared with and without a geomagnetic sensor. On this simulation, in order to make clear the effect of a geomagnetic sensor fusion, no noise and bias are included to the sensor observables, and sensor biases are not estimated. Information of a geomagnetic sensor, which is a tri-axis vector, is incorporated to the observed values of the EKF without changing its scheme. As results, although in the “without” case,?errors of attitude angles are in the order of 0.1 degrees, in the “with” case the errors decrease in the order of 0.01 degrees. In addition, against that a heading error does not converge until over 1000 seconds in the previous, the error immediately converges in the following. These results show that the algorithm is effective in improving the attitude accuracy. It is proved to be effective to utilize a tri-axis vector obtained from a geomagnetic sensor without changing its scheme and incorporate it into the EKF. In the last, the algorithm is validated with an actual aircraft experiment. The system using the proposed algorithm keeps high precision and reliability during the flight including all motions of an aircraft.
Published in:	Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013) September 16 - 20, 2013 Nashville Convention Center, Nashville, Tennessee Nashville, TN
Pages:	1071 - 1083
Cite this article:	Kosaka, Y., Tsuchiya, T., Naruoka, M., Tomita, H., Kurihara, H., Ishida, K., Ichikawa, T., "Development of High-precision Navigation Algorithm by Fusion of GPS and MEMS Sensors," Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013), Nashville, TN, September 2013, pp. 1071-1083.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In