- 學位論文
多感測器資料融合與自走車導引之研究
The Research of Multisensor Data Fusion and Mobile Robot Navigation
摘要
本篇論文提出一個T-S模糊模型為基礎的平行分佈訊息濾波器，做為GPS/INS訊息整合的軸心。首先，對於全球衛星定位系統（GPS）和慣性導航系統（INS）的基本概念稍作介紹。由於兩者之間的互補性，因此產生了GPS/INS融合的機制。卡門濾波器是一非常有用的估測工具，藉由卡門濾波器的基本觀念可以推導出訊息濾波器。此濾波器相較於卡門濾波器而言有運算量較少的優點。接著，提出T-S模糊平行分佈訊息濾波器。第一步，將一非線性系統以T-S模糊模型表示成數個線性次系統。第二步，利用訊息濾波器估測每一個線性次系統的狀態。最後，利用模糊推論輸出，便可以獲得原非線性系統的狀態估測值。另外，分別利用平行分佈訊息濾波器和平行分佈卡門濾波器對自走車作狀態估測的模擬。實驗部分，以本研究室之自走車結合平行分佈訊息濾波器實現資料融合的目的。
並列摘要
This thesis presents a Takagi-Sugeno fuzzy model-based parallel distributed Information filter for GPS/INS sensor fusion. Due to the complementary, the sensor fusion of GPS and INS has widly potential application in navigation. Compared to Kalman filter, the algebraic equivalent filter -- Information filter is introduced and the advantage of it is discussed. Based on T-S fuzzy model, we propose a T-S fuzzy parallel distributed Information filter to estimate the state variables. First, we need to model a nonlinear system into T-S fuzzy model, which consists of local linear subsystem . Second, information filters are applied to each linear subsystem. Using the fuzzy inferred output, the overall estimate of state for the original nonlinear system is obtained. The comparison of numerical simulations on the mobile robot by parallel distributed Information filter and parallel distributed Kalman filter is illustrated. Finally, an experiment of data fusion using the parallel distributed Information filter is implemented.