| 摘要: | 本論文提出一個以六個紅外線感測器為基的避障方法與一個以視覺為基的射門方法，並且將其實現在人形機器人上。本論文首先對所設計與實現的小型人形機器人做一個整體的描述，機器人的高度與重量分別為43公分及3.4公斤。26個自由度的機構設計讓機器人可以完成前進、左右轉、左右平移、尋球與射門的動作，自由度的規劃分別在腳部有14個自由度、在腰部有2個自由度、在手部有8個自由度以及在頭部有2個自由度。本論文使用CMOS感測器、紅外線感測器與電子指南針來讓機器人可以獲得環境資訊。本論文以Nios II發展板做機器人的運動控制器，而十六位元數位訊號處理器μ''nsp負責處理CMOS感測器所擷取的物體影像資訊以及決定機器人的動作行為。本論文以Borland C++ Builder程式架構一個人機介面來讓使用者可以容易的設計與規劃機器人的動作。在避障的設計與實現上，本論文依據電子指南針與紅外線感測器分別獲得機器人與目的地的方向資訊以及機器人與障礙物的距離資訊，提出一個可以在五種動作(左平移、左轉、前進、右轉、右平移)當中決定一種最適合的動作來閃避障礙物的方法。從連續圖示所展現的實驗結果中，可以看出所提的避障方法確實可以讓機器人能夠自主閃避障礙物，並且有效的往目的地方向移動。此外，在射門的設計與實現上，本論文依據機器人頭上之CMOS感測器所擷取的影像資訊，提出一個視覺為基射門的方法。由實際測試中，可以看出所提的方法確實可以讓機器人能夠自主的尋球及決定移動路徑，並且移動至一個適當位置踢球。
In this thesis, an obstacle avoidance method based on six infrared (IR) sensors and a vision-based shoot ball method are proposed and implemented on a humanoid robot. First, a design and implementation method of a small-size humanoid robot with 26 DOFs (degrees of freedom) is described. The height and weight of the implemented robot is 43 cm and 3.4 kg, respectively. A mechanical structure is design so that the implemented humanoid robot can walk forward, turn right and left, slip right and left, find the ball, and kick the ball. There are 2 DOFs on the head, 2 DOFs on the trunk, 4 DOFs on each arm, and 7 DOFs on each leg. A CMOS sensor, six infrared sensors, and a digital compass are integrated so that the robot can obtain the environmental information. A Nios II development board is used to be the motion controller of the robot. A 16-bit DSP processor, μ’nsp, is used to process the image captured by a CMOS sensor to find some objects and decide an action behavior. A human-machine interface established by Borland C++ Builder is constructed to let the user can easily design and plan the behaviors of robot. One electronic compass and six infrared sensors are mounted on the robot to obtain the head direction of the robot and detect obstacles, respectively. Then, based on the obtained information from these sensors in each decision, an obstacle avoidance method is proposed to decide one behavior from five movements: walk forward, turn right and left, and slip right and left. One real experiment is presented to illustrate that the robot can autonomously avoid obstacles and go to the destination effectively. Furthermore, a vision-based shoot ball method is proposed based on the image information captured by the CMOS sensor installed in the head of the humanoid robot. From the practical test, we can see that the implemented robot can autonomously find the ball, determine a mobile path, and walk to the desired position to kick the ball. Some experimental results are presented to illustrate that the implemented humanoid robot can autonomously avoid obstacles and shoot ball. |
