Study on Monte Carlo Localization Based on Omnidirectional Vision Systems

[[abstract]]移動機器人定位是目前非常熱門的問題，各類研究都對定位有不同解決方式，而Monte Carlo定位則是目前最為流行的定位法，因為有著不需要初值，以及易於實現的優點。因此本論文致力於其定位方法的探討，由於各家機器人在針對sensor model 感測部份，大部分都採用昂貴的雷射掃瞄儀或是超音波感測器，而本文則利用廉價的凸面鏡來仿效雷射掃瞄儀的效果並且利用電子羅盤來估測機器人朝向角，實驗結果說明亦具有不錯的定位效果。而為詳細探討Monte Carlo 的定位技術，本研究亦發展一模擬系統，以便探討各種參數對Monte Carlo 定位效能的影響。探討內容包括掃描射線個數與粒子數對定位誤差的影響，實驗包括移動機器人在球場中可以快速自我定位。[[abstract]]Mobile Robot Localization is popular in recent years Each kind of research all has different way to solve the localization but Monte Carlo Localization (MCL) is almost the most popular localization because does not need initial position and has the advantage of easy realization Therefore this research devotes to studying MCL Various robots locate its position based on expensive sensors such as laser rangefinder or ultrasonic In this thesis we use the inexpensive convex mirror to imitate the effect of laser rangefinder and estimate robot heading with the electronic compass The experimental results demonstrate the well performance of MCL In addition the MCL technology is discussed This research also develops a simulation system in order to explore Monte Carlo localization influenced by its parameters The exploration includes the localization position error influenced by the numbers of scanner lines and particles respectively The experiments also include that the robot is able to finish fast self- localization in the workspac