本研究探討了一種適用於小型無人機的避障演算法FlowDep，此種演算法的靈感來自果蠅或蜜蜂這樣的昆蟲，基於光流來估算深度，僅使用無人機相機以及位姿資訊，不需要額外的感測器或計算資源。因此適用於資源有限的無人機平台。使用了Gazebo模擬無人機飛行來測試該演算法的效能。本研究透過Gazebo建立了一個樹林場地，模擬無人機穿過樹林環境，並統計是否能成功穿越而不撞上障礙物。在調整不同參數下共進行了300次模擬測試，成功率介於46~92%，同時評估了光流、相機解析度、以及動作控制等參數對成功率的影響。這些結果顯示，FlowDep演算法具有實際應用價值，可為低成本無人機提供一種新的避障方案。

This study explores an obstacle avoidance algorithm, FlowDep, for small unmanned aerial vehicles (UAVs). The algorithm is inspired by insects such as fruit flies and honeybees, and it estimates depth using optical flow, relying solely on the UAV's camera and pose information without the need for additional sensors or computational resources. This makes it suitable for resource-constrained UAV platforms. The performance of the algorithm was tested by simulating UAV flights in Gazebo. A forest environment was modeled in Gazebo to simulate the UAV's traversal through the forest, and success rates in avoiding obstacles were statistically analyzed. A total of 300 simulation tests were conducted under different parameter settings, yielding success rates between 46% and 92%. The study also evaluated the impact of parameters such as optical flow, camera resolution, and motion control on the success rate. The results demonstrate that the FlowDep algorithm has practical application value and provides a new obstacle avoidance solution for low-cost UAVs.