一個針對行人位置推估系統的速度校正方法

Abstract

隨著微機電技術的日新月異，各式低價、體積又小的慣性感應器已大量出現。使 用慣性感應器應用於行人導航上已在最近引起各方的注意。在行人慣性導航的方法中，對慣性感應器所測到的加速度作二次積分，來推估位置是最直接有效的方法，但前提是需能夠克服因感測器精度不足，加速度在對時間積分後快速產生的累計誤差，在過去可將感測器安裝在足部利用ZUPT 的技巧，在行走過程中足部著地時速度大小應為零的特性，來校正行人導航的速度，但此方法限制是需將感測器安裝在足部，而行走過程中足部運動的變化非常劇烈，反而有感測器量測誤差擴大的副作用，特別是前進方向上的量測上。現在本研究提出另外一種累計誤差的校正方法，並可避免量測誤差的擴大，該方法稱作Walking velocity update 簡稱WUPT，該方法使用2 組感測器，一組安裝在軀幹上，因上半身的運動變化較不劇烈，可量測到較可靠的前進方向和加速度，使PDR 系統進而推估出較準確的行人速度和位移，另一組sensor 安裝在小腿上量測角速度，於行進中該腿垂直站立於地，大小腿幾無相對運動的時刻，透過牛頓運動定律反推上半身 應有的正確速度，來校正系統的推估速度。我們已利用該方法實作出新型的行人慣性導航系統，經驗證該系統的定位性能的確優於ZUPT PDR 系統。

Along with the MEMS technology development, there are various low-cost and small-volume inertial sensors produced massively. Using inertial sensors for Pedestrian Dead-reckoning has attracted a lot of attention recently. It is the most efficient and direct way to estimate the displacement by using the integration of acceleration, but the prerequisite is to overcome the increase of the accumulated error with time rapidly. In the past, there was a famous ZUPT technique, whereby the researchers installed the sensor on the pedestrian’s foot and corrected the speed of the PDR system by using the characteristic, which is that the speed of the foot is zero while the foot is landing on the ground during the movement. The limit of this method is that the sensor has to be installed on the foot. The side effect of the sensor placement on the foot is that the dramatic movement would extend the sensor measurement error, especially for the heading determination. The contribution of this thesis is to propose another kind of correction method for the accumulated error and avoid the disadvantage of the ZUPT technique. This kind of correction method is called the WUPT technique, short for the Walking Velocity Update. The technique uses two sets of the inertial sensors on the pedestrian. One set is installed on the torso due to more stability of the movement, so we can obtain a more accurate measurement of the heading direction and the acceleration. The other set is installed on the pedestrian’s low leg to measure the angular rate. We can correct the estimate speed of the torso with the angular rate of the low leg at the moment the thigh and the low leg become a straight line vertical to the ground during the walking. We have used this method to make the WUPT PDR system prototype and have verified that its localization performance issuperior to the ZUPT PDR system.