即時視覺輔助戶外無人車系統設計與控制

Abstract

本篇論文主旨為發展戶外自動導航車系統，使導航車能追尋路面預先規劃好之標線軌跡，藉由全球定位系統(GPS) 做全域定位，影像系統偵測路面標線做區域控制，且建立由無線網路構成的各次系統資訊交換機制，使主控站得以監控載具行為。在無人載具追蹤標線控制方面，為使追蹤目標更明確，採用黑白相間的標線來增強對比度，利用樣板相關函數判斷影像中是否存在標線，再經由邊緣偵測、隨機霍氏轉換計算出標線參數。控制方面則採用模糊控制理論，根據標線資訊，經由知識庫推論載具所需要的速度輸出。若載具視覺系統的初始狀態並未偵測到標線，則利用GPS的資訊，控制載具往標線方向前進。為提升GPS定位精度，採用電碼差分定位法(DGPS)，其精度可達兩公尺。當載具視覺系統只偵測到單一標線時，則此時控制權在視覺模式；若偵測到一條標線以上時，即代表載具位於十字路口，則此時控制權切換至GPS模式，以控制載具姿態，使其朝目的地前進。本篇論文最後以具體實驗來驗證本系統之可行性。

The main theme of this thesis is to develop an integrated navigation and control system for an autonomous vehicle using Vision and DGPS in outdoor environments. The vehicle uses vision to detect landmarkers and DGPS information to determine the vehicle's position. The vehicle is constructed step by step and the data exchange mechanism between sub-systems is set up through wireless network. The task of path following is performed such that the vehicle can track along a landmarker specified in advance. In the path following control problem, we use the Fuzzy control theory to determine the velocity of the vehicle. We use black-and-white landmarks in order to increase the contrast between the roads and the landmarks. The concept of template correlation is used to identify existing landmarkers in vision. The techniques of Edge Detention and Randomized Hough Transform are then applied to obtain the parameters of the landmarker. If no landmarker is detected by the vision system, the DGPS is used to control the vehicle approaching the specified position of the landmarkers. When the vehicle is passing a crossroad, we also use the GPS information to control the attitude of the vehicle toward the destination. Experimental results showed the effectiveness of our proposed navigation and control methodology.