- 學位論文
基於影像辨識和類神經模糊方法之避障機器人
An Obstacles Avoidance Method Based on Image Recognition and Neuro-Fuzzy Theory for a Mobile Robot
摘要
在本論文中，我們提出了移動機器人基於影像辨識和類神經模糊理論的避障方法，我們主要的目的在於希望機器人能藉由這個系統，避開凸起和凹陷的障礙物，以增加機器人移動時的安全性。 在本論文中，我們會介紹機器人避障的流程，首先是先介紹機器人的模型，我們會裝置三個超音波感測器和一個CCD的鏡頭，我們利用這些感測器來偵測障礙物。第二步，我們設計一個類神經模糊的系統，我們會建立它的隸屬函數和模糊規則，然後訓練和修正這些參數，使這個系統更加完善，第三步，我們介紹避障的過程，超音波感測器會偵測凸起的障礙物，我們可以得到機器人與他們之間的距離，CCD鏡頭則會藉由影像辨識技術偵測凹陷的障礙物，例如坑洞或窪地。這些數據將會成為類神經模糊系統的輸入，經過系統的計算，我們可以獲得輸出，我們用輸出結果來改變機器人移動的方向，避開這些障礙物。 最後，我們會用實驗來證明這個系統的可行性。我們利用Matlab來進行實驗的模擬，我們設計一個有凹陷障礙物的迷宮，我們計算輸入的資料，並且經由避障系統得到輸出。實驗結果顯示，機器人能成功地避開凸起和凹陷障礙物即便這是個複雜的環境。我們確實提升了機器人移動上的安全性 本論文中我們的研究有下列幾點貢獻: (1) 安全性:我們的系統能夠偵測凹陷的障礙物並且避開它，使避障系統更為安全且完整。 (2) 適應性:我們考慮了更多狀況，並設計模糊規則庫，可以使機器人適應更多複雜的環境 (3) 增加用途:由於我們能成功地避開凹陷的障礙物，機器人能夠完成更多任務，例如探勘月球之類的，我們確實地增加了機器人的用途。
並列摘要
Abstract In this thesis, we propose an obstacle avoidance robot based on Image Recognition and Neural-Fuzzy control systems. The main purpose in this thesis is to avoid concave and convex obstacles by this system, to increase the safety when the robot is moving. In this thesis, we will introduce the process of obstacles avoidance. First, we describe a mobile robot. There are three ultrasonic sensors and a CCD camera on it. We use these sensors to detect the obstacles. Second, we design a Neural-Fuzzy system. And we create the membership function and fuzzy rules. Then, we train and correct the parameters to make system better. Third, we will introduce the procedure of obstacles avoidance. We detect obstacles by sensors. The ultrasonic sensors detect the convex obstacles and then we can obtain the distance between robot and obstacles. The CCD camera detects the concave obstacles like a pit or a hole by image recognition technology. These data we get will become the inputs of Neural-Fuzzy system. By calculating, we can get the output. We use the output to control motors to change the direction of moving. And we can avoid the obstacles. Final, we prove the feasibility of this system by experiment. We simulate the experiment by Matlab. We create a maze with concave obstacles, and then we calculate the input data and get the outputs by the obstacles avoidance system. The experimental results show that Robot can avoid obstacles successful even it is a complex environment. It actually increases the safety when robot is moving. In this thesis the contribution of our research are as follows: (1) Safely: In our system, we can identification concave obstacles and avoid it. It makes obstacles avoidance system more safe and complete. (2) Adaptability: We consider more situations and design fuzzy rules. This make the obstacles avoidance system can adapt in more complex environments (3) Ability: Because we can avoid concave obstacles successful, the robot can do more tasks, like explore on moon. It increases the ability of mobile robot.