以DSP實現GPS與IMU感測器整合系統

Abstract

本文目的在發展一套以Extended Kalman Filter作為演算法之基礎架構，整合GPS與IMU的訊號用以提高運動物體的軌跡與姿態判定的精準度。GPS可以提供一運動物體於3D空間中的位置與速度，然而其輸出頻率僅有1Hz，定位精度約10公尺。IMU包括一3軸的加速規與3軸的陀螺儀，可以提供三軸的加速度量測與三軸的角速度量測。其輸出頻率可高達1KHz，然而其輸出訊號亦包含雜訊且易出現訊號飄移（drifting）。因此若能整合GPS與IMU的輸出，將有機會獲得一高輸出頻寬、高精度的6DOF定位資訊。 本文選用Extended Kalman Filter來整合GPS與IMU輸出的目的在於：(1)它可以整合具不同輸出頻率的感測器；（2）它可藉由其訊號整合方式、運算方式來消除各個感測器量測雜訊（noise）對整體訊號估測的影響，提高定位精度。至於訊號飄移（drifting）的部分，本文嘗試建立多種的GPS與IMU整合架構，並探討如何利用不同的整合架構來消除IMU訊號漂移所造成的估測誤差，其研究結果列表於本文的結論一章中。 最後，本文介紹IMU訊號輸出的校正方式，為此一感測器整合系統準備。並以一DSP平台來實現此一訊號整合系統。目前，IMU輸出的類比訊號與GPS輸出的NMEA 0183之ASCII Code數位訊號，皆可同步的經由DSP來接收。以C-code來實現Extended Kalman Filter的工作仍在持續中。

This thesis aims to develop a sensor fusion system and implement it on a DSP platform. This sensor fusion system uses Extend Kakman filter to coordinate GPS output and IMU output to obtain accurate trajectory and attitude for an object in motion. The GPS can provide the position and velocity information. However, its data output rate is 1 Hz and the position accuracy is around 10 meter. The IMU, consisting of 3-axis accelerometers and 3-axis gyroscopes, can provide 3-axis accelerations and 3-axis angular rate measurements with a data rate around 1KHz. However, its signal is noisy and drifting. Therefore, it is preferred to combine these two sensors to obtain a high data-throughput and high accurate 6 DOF sensor fusion system. The advantages of using Extended Kalman filter are that it can coordinate sensors with different data output rate and it can minimize the effect from measurement noise associated with each sensor. As for the signal drifting from IMU, this thesis discusses different sensor fusion systems that comprised of different sensor units. This research result is summarized in table shown in the conclusion section in this thesis. Lastly, this thesis describes the detail procedures of sensor calibration which is needed prior to the construction of a sensor fusion system. This sensor fusion system will be implemented on a DSP platform. Currently, the GPS data, which is NMEA 0813 in ASCII codes, and the IMU data, which is analog, can feed in the DSP synchronously. The C-code coding for the Extended Kalman filter is on the way.