導盲機器人之路徑規劃與控制

Abstract

這篇論文介紹一個導盲機器人的研發，其設計之目的在幫助盲人或有視覺傷害的人，能安全的繞過障礙物或突然出現的物體，而到達他所要去的地方。我們首先發展一套法則來做路徑規畫，這個路徑規畫方法可以找出最短的路徑。藉由最佳控制設計發展一個路徑追蹤控制器，讓機器人可以跟隨一個預設的軌跡。為了以電腦視覺求得障礙物的位置和方向，必須利用CCD的內外參數來做校正，並以校正的的結果作影像分群、匹配和3D位置計算來得到障礙物的位置和大小。並藉由分類網路的設計，可以快速的作區域避障的路徑規畫。使用者透過手握的桿子來感覺驅動輪改變命令的物理作用力來跟隨導盲機器人。

In this thesis a guidance robot is designed to help blind or visually impaired pedestrians to navigation safely among obstacles and other hazards faced by blind pedestrians. We develop a path planning method to find a shortest path. An optimal controller has been design to perform the path tracking task in order to make the guidance robot to follow the predefined path. In order to obtain absolute orientation and position using computer vision system, each individual camera needs accurate calibration to acquire internal and external parameters. The calibration results are used in the clustering, stereo matching and 3D position calculation to find obstacle's position and width. The mobile robot will make fast local path planning using a hueristic clustering network to prevent from collision with unexpected obstacles on its pre-defined path. The user can feels the steering command as a very noticeable physical force through the handle and be able to follow the guidance robot's path easily without any conscious effort。