本論文將完成視覺定位及建圖(SLAM:Simultaneous Localization And Mapping)於二輪式移動式機器人，並將擴張型卡爾曼濾波器(EKF:Extended Kalman Filter)建立於現場可程式化閘陣列系統(FPGA: Field Programmable Gate Array)上。首先以機器人上之雙鏡頭感測器偵測立體影像，再使用SURF(Speeded-UP Robust Features)搜尋立體影像上之影像靜態地標，地標透過EKF估測機器人及地標的大地空間位置，最後利用模糊控制器完成移動式機器人控制，使移動式機器人可穩定行進於空間中並建立空間地圖。本論文成功簡化EKF運算量，並將其實現於FPGA上，同時將二輪式移動機器人運動學運用在EKF演算法中。最後本論文於實驗中證實簡化EKF之成效及SLAM系統之準確性。 For the issue of an autonomous mobile robot, the capabilities of localization and mapping are very important. In this paper, a visual simultaneous localization and mapping (SLAM) of a two-wheeled mobile robot (TWMR) is established by the extended Kalman filter (EKF) and speeded-up robust features (SURF). First, the stereo images are captured by a binocular vision system build on the mobile robot. Then the interest landmarks of an image are detected by SURF. The geometric relationship between landmarks and robot are derived by the kinematics of TWMR and EKF. The motion control of TWMR is based on fuzzy control design. We also reduce the computation demand of EKF and successfully implement the simplified EKF on the field-programmable gate arrays (FPGA) architecture. Finally, the practical experiments prove efficiency of the modified EKF and accuracy of the developed SLAM system.
