本研究旨在整合GPS接收機、電子羅盤、雙眼相機、超音波感測器及動力系統,以發展一套戶外自動導航車系統。該系統包括主控站、參考站及無人載具,無人載具和參考站以GPS訊號進行載波平滑電碼法定位獲得載具位置,無人載具搭載的電子羅盤感測載具目前方位,雙眼相機及超音波感測器負責偵測障礙物,若遭遇障礙物則進行避障,主控站負責監控載具的行為,並以無線網絡和載具交換資訊。 雙眼相機系統部分,本論文採用SRI Stereo Engine的軟體開發介面(API),搭配網格化的概念判定障礙物的位置,並應用快速探索隨機樹規劃避障路徑,避開雙眼相機偵測到的障礙物。超音波感測器部分,由FPGA開發板控制六顆超音波感測器的輸入輸出,並利用規則庫系統針對當下感測到的障礙物資訊推論出載具應有的速度輸出。 針對上述兩項避障模式,本論文設計了一套整合流程,使載具能視當下狀況選擇合適的避障模式,並能在兩模式間來回切換,以適應不同環境。本論文設計了數個實驗驗證各感測器及整合流程的可行性,證明本論文所提出的策略確實可行。
The main purpose of this paper is to integrate GPS recievers, electronical compass, stereo camera, ultrasonic sensors, and dynamic system, developing a PC-based unmanned vehicle system for outdoor navigation. The main system comprises three parts: main station, reference station, and the vehicle. GPS signals received by reference station and vehicle are utilized to determine the position of the vehicle by Carrier Smoothed Code method. Electronical compass senses the heading angle of vehicle. Stereo camera and ultrasonic sensors are employed to detect obstalces, and obstalce-avoiding action is taken if obstacles exist. Main station monitors the motion of the vehicle, and the information is exchanged via wireless network. SRI Stereo Engine(API) and the concept of “grids” are utilized to determine the relative distance between the vehicle and obstalces. Once the positions of obstacles are known, a path along which the vehicle can avoid those obstacles is decided by Rapidly-Exploring Random Tree method. Six ultrasonic sensors are assembled and controlled through FPGA board, and a rule-based system is utilized to deal with the data collected by ultasonic sensor system, determining the corresponding linear velocity and angular velocity output in order to avoid nearby obstacles. An operation flow is developed to integrate these obstacle avoidance modes, making the vehicle choose the proper mode. Besides, the vehicle is able to switch from one mode to another correctly. Experimental results described in this thesis proves that the proposed navigation and control methodolgy are feasible. Keywords: Rapidly-Exploring Random Tree, Obstacle Avoidance for Navigation System.