本文旨在建立精準的相對定位系統，在使用者設定初始位置後，藉由自走車每次動作的距離及方向角推算自走車位置。其中自走車選用iRobot公司出產之可編程機器人(簡稱iRobot)，距離資料由iRobot內建里程計取得。經實驗後，發現iRobot內建馬達編碼器會依每次的轉動而累積誤差，且無法一次轉動超過90度。所以本系統加裝意法半導體公司的iNEMO感測模組，取用陀螺儀z軸及磁力計的x、y軸資訊來修正目前自走車的方向角。 其中角度修正分為三部分，首先經由累加計算及換算求出陀螺儀轉動值，再將取得之磁力計值通過中值濾波器得知兩筆角度資訊；其二討論iNEMO感測模組的誤差特性並將之用模糊規則描述以預測真實角度；最後，用從實驗得知的誤差範圍設計一組卡爾曼濾波器去除非規則性雜訊，得到精準之方向角。為了證實方法的可行性，提出指定位置追蹤法，討論在不同周期加入角度修正(每兩指定位置距離或一段短距離)的差異。其中在每兩指定位置距離週期下，額外在角度修正較準確的模糊補償及卡爾曼濾波後加上立即補償進行比較。 本文進行三階段實驗，首先進行「轉角實驗」，收集陀螺儀及磁力計之誤差趨勢；接下來進行「角度校正的模擬值分析」，模擬演算法的修正效果；最後進行「指定位置追蹤實驗」，收集其指定路徑、演算路徑及真實路徑的比較，並進行綜合比較探討所有方法的效能及誤差。 In this thesis, we provide a precise relative positioning system. After the user sets the initial position, positions of a mobile robot are calculated by action of a mobile robot every time including distance and direction. We select iRobot Create? Programmable Robot (called iRobot) to be a mobile robot, and data of distance is obtained by a built-in odometer of iRobot. In experiment, we notice that there is cumulative error of direc-tion each time iRobot rotating by a built-in motor encoder, and it cannot be rotated more than 90 degrees once. Therefore, iNEMO module of STMicroelectronics is installed in this system. The information of z-axis gyroscope and x, y-axis magnetometer are used to correct the current direction of a mobile robot. Angle correction is divided into three parts, first we calculate direction with data of gyroscope and magnetometer second, we discuss the error characteristics of iNEMO module, and describe them by fuzzy rules to predict the real angle finally, we design a Kalman filter to remove irregular noise, and get the precise direction. In order to confirm the feasibility of methods, we design a specified location of tracing method, and discuss different cycles of angular correction (a distance of two specified location or a short distance). In the cycle of a distance of two specified location, we compare additional immediate compensation after using more accurate angle correction such as fuzzy and Kalman filter.
