以全方位視覺之特徵掃描比對方法進行機器人自我定位

Robot Self-Localization using Feature Scan Matching with an Omni-directional Vision

陳敬宏(Jing-Hong Chen)

指導教授 ： 王銀添

淡江大學/工學院/機械與機電工程學系碩士班/碩士(2007年)

https://doi.org/10.6846/TKU.2007.00892

摘要

本論文規劃足球機器人的蒙地卡羅自我定位方法，所使用的機器人具備全方位視覺與全方位驅動輪。運動模型方面，以全方位驅動裝置的反向運動矩陣相對時間積分，搭配回授的里程計訊息建立運動模型。感測模型方面，以機器人位置為中心，規劃全方位影像徑向的像素掃描線。將像素掃描線偵測到的特徵，以高斯分佈方式建立感測模型的比對資料庫。機器人行進中執行自我定位時，交互使用運動模型與感測模型修正機器人在環境中的位置信念(position belief)。

關鍵字

全方位視覺 ； 蒙地卡羅自我定位方法

並列摘要

In this thesis, we use Monte Carlo localization (MCL) algorithms to solve the self-localization problem for robots and apply to soccer robots which have an omni-directional vision system and an omni-directional-driven mechanism. The differential kinematics equation for the omni-directional drive is derived to construct the motion model of MCL. In the motion model, an optical encoder is utilized as the odometer sensor. For the sensor model of MCL, an omni-directional vision system is mounted on the center of the robot to detect color features of the environment. The database of the color features is built for the feature scan matching by adopting the method of mixture probability distribution. After the MCL algorithms are developed, the robot system can locate its position in the environment by updating its position belief recursively, according to the motion model and the sensor model of MCL algorithms.

並列關鍵字

omni-directional vision ； Monte Carlo localization

參考文獻

陳子原，2006，自主足球機器人系統之行為式控制，淡江大學機械工程學系碩士論文。

J. Aranda, A. Grua, and J. Climent, 1998. “Control Architecture for a Three- wheeled Roller Robot”, Proceedings International Workshop on Advanced Motion Control.

K. Arras and S. Vestli, 1998. “Hybrid, high-precision localization for the mail distributing mobile robot system MOPS”, In Proc. of the Int. Conf. on Robotics and Automation, 1998.

S.A. Bazaz and B. Tondu, 1998. “3-Cubic Spline for On-Line Cartesian Space Trajectory Planning of an Industrial Manipulator”, IEEE AMC ‘98-COIMBRA, pp. 493-498.

W. Burgard, D. Fox, B. Cremers, 1998. “Integrating Global Position Estimation and Position Tracking for Mobile Robots：The Dynamic Markov Localization Approach”, In Proc. IROS Conf., Victoria,BC, Canada, Oct. 1998, pp. 730–735.