本論文提出一個結合粒子群最佳化法（PSO）之雙層粒子濾波器架構，並將其應用於移動機器人的定位與地圖建置。在機器人定位的部分，本論文提出一個結合粒子群最佳化法之粒子濾波器架構，透過粒子群演算機制所具備的快速收斂與強大的最佳解搜尋能力等優勢，可增進移動機器人定位的性能。相較於其他的定位方法，本論文所提出的架構能夠更精確的估測得到移動機器人在環境中的位置座標。在地圖建置的部分，本論文以粒子濾波器對環境中的特徵物體進行位置估測，並將粒子濾波器內部的預測機制予以改良，解決特徵物體在估測時無預測資訊輸入的問題，並使粒子具有小幅度的擾動，此改良型的粒子濾波器可提升粒子濾波器在進行地圖建置時估測的正確性。最後將兩者予以整合，建立結合粒子群最佳化法之雙層粒子濾波器於移動機器人定位與地圖建置的系統架構，藉由演化機制的運作可以改善移動機器人定位的性能並提升地圖建置的正確性，實現移動機器人在探索與認知未知環境的能力。模擬與實驗的結果證實，本論文所提出結合粒子群最佳化法之雙層粒子濾波器具有不錯的性能，能夠滿足移動機器人探索未知環境的應用需求。 In this dissertation, architecture of two-layer particle filters incorporating particle swarm optimization (PSO) is proposed and applied on the localization and mapping of mobile robot. For the robot localization, the particle filter is modified by integrating a particle swarm optimization algorithm, where the excellent performance in global optimization of the PSO is used to improve the localization performance. In comparison with conventional particle filters, the proposed particle filter can better determine the robot’s position. For the map building, the particle filter is applied to estimate position of landmarks in the environment, in which the prediction step in the filter is modified by adding small random perturbations into the particles. As a result, the proposed method can better determine position of landmarks. By combining these two functionalities, the architecture of two-layer particle filters is proposed to investigate the localization and mapping of the mobile robot simultaneously. Due to the incorporation of the PSO, the proposed architecture is capable of reducing the localization error of the robot while improving the mapping accuracy of the landmarks. As a result, the robot can better explore an unknown environment with the proposed architecture. Simulation and experimental results show that the proposed approach has a better performance for the localization and mapping of the mobile robot.
