使用PSO調整之增強型ICP演算法於未知環境地圖之建立

Abstract

本論文使用Pioneer 3-DX兩輪自走車搭載一台LMS-100雷射測距儀做未知環境的地圖建置，主要使用ICP演算法將每一筆雷射測距儀的掃描資訊疊合，但由於傳統ICP演算法本身容易受到雜訊與離散點影響，造成配對到不恰當的配對點，產生對齊有誤差，而在雷射掃描儀的連續掃描下，誤差的累積越來越多，導致整體的環境地圖對齊結果並不理想，故本論文提出使用PSO調整增強型ICP演算法來克服其問題，先使用PSO演算法將要對齊的兩集合做初步的對齊，避免兩集合落差太大產生區域最佳解，接著使用部分全域的地圖當作參考資訊，搭配篩選重疊資訊模組、權重模組及參考地圖間格模組，成為增強型ICP演算法，此演算法不但可以克服雜訊與離散點影響，還可以降低配對到不恰當的配對點，增加對齊效果，降低累積誤差，以獲得更佳的未知環境地圖。This paper proposes a PSO-tuned enhanced iterative closest point algorithm (ICP) to build maps for an unknown environment using a Pioneer 3-DX wheeled mobile robot with a laser measure scanner (LMS-100). The proposed algorithm first aligns each scanned information by the ICP algorithm. Because traditional ICP algorithms are easily affected by noise and outliers, false matching points and alignment errors are therefore inevitable. As a result, there are more and more errors accumulated as the scanning process by the laser scanner continues, which results in imperfect alignment of the environmental map as a whole. Therefore, this paper proposes the use of Particle Swarm Optimization (PSO) to work with the Enhanced-ICP in order to effectively filter out outliers and avoid false matching points during the map building process, where PSO is used to align two data sets to avoid huge transformation that causes local optima. Then, we use part of global map as the reference data set with overlapping points for subsequent data matching. The proposed algorithm not only improves outlier and noise problem but also reduces false matching points so that it has better alignment and smaller accumulated errors. As a result, good environmental map is obtained.