本論文提出一套改良式SLAM系統，利用雷射測距儀取得實際環境資訊，此資訊可被視為環境物體幾何特徵資訊，然後將此特徵資訊應用在同時定位與地圖建置(SLAM)問題。在環境曲率特徵擷取模組上，本文採用改進K-Value演算法抓取環境特徵，將感測資訊視為一包含環境物體幾何特徵資訊之曲率線段，然後透過識別出基於不連續面所產生的斷點，並計算兩斷點間線段的曲率值，將高於門檻值之環境曲率特徵，將其視為環境地標群。在同時定位與地圖建置模組上，本文提出一套改良式FastSLAM演算法，以FastSLAM演算法做為架構，著重於加快運算速度，經過前處理後，利用高可信地標群之三角定位法更新機器人狀態，如此可維持地圖建置之精確率。經過模擬與實驗驗證，本文所提出之改良式SLAM系統，在於移動機器人之同時定位與地圖建置上，具有較佳的性能與提昇執行效率。 In this thesis, a modified SLAM system is proposed. This system use a laser range finder to accesses to environmental information, which represents the geometric features of objects in the environment and then it is applied on simultaneous localization and mapping (SLAM) problem. In the environmental curvature features extraction module, capturing environmental features use an improved adaptive K-Value algorithm. Sensor data is considered a curve segment which included geometric features and then extracted breakpoints based on discontinuous plane. Computing a curvature of a line segment between two breakpoints is considered environmental landmarks when the curvature over the threshold. In the SLAM module, a proposed modified FastSLAM algorithm is based on FastSLAM algorithm, it focuses on improving computing speed. After pre-processing, updating robot status by high reliability landmark swarm triangulation method can be maintained accurately on mapping. Simulation and experimental results show that the proposed modified SLAM system has a better performance and enhancing efficiency for the SLAM of the mobile robot.
