本論文建構了一個適用於無人載具的路徑追蹤與導航避障策略 ; 在研究中使用位於天頂的廣角網路攝影機來取得環境資訊及載具與動態障礙物資訊,接著利用位能場快速搜索隨機樹(Potential Field RRT*)演算法規劃靜態避障路徑,並分別透過模糊控制器與類神經模糊控制器控制載具追蹤路徑,同時藉由攝影機隨時監測動態障礙物以進行路徑微調避免碰撞。 模擬與實驗結果顯示,本研究發展之路徑規劃演算法能夠有效在靜態環境中規劃出一條遠離障礙物的路徑,且於環境中存在動態障礙物的情況下仍能即時調整路徑以避開障礙物。在控制器設計方面,類神經模糊控制器相較於傳統模糊控制器更能使載具確實貼合設定路徑行走,在實際導航任務上具有較高應用價值。
This paper constructs a path tracking and obstacle avoidance strategy which suitable for unmanned vehicles. In the research, a wide-angle camera located at the zenith is used to obtain environmental information, vehicle and dynamic obstacle information. The Potential Field RRT* method is then applied to design the optimal path to avoid static obstacles. To track the designed path, the fuzzy controller and the neuro-fuzzy controller are implemented separately. In the mean time, the camera monitors the dynamic obstacles at any time to fine-tune the path to avoid collision. Simulation and experimental results show that the path planning algorithm developed in this study can effectively plan a path away from obstacles in a static environment, and can adjust the path in real time to reduce the risk of collisions with moving obstacles. In terms of controller design, it is seen that the neuro-fuzzy controller can make the vehicle follow the set path more accurately than the traditional fuzzy controller, and thus has higher application value in practical navigation tasks.