飛行機器人於溫室資訊智慧偵測系統之整合研究

Abstract

溫室內的設施栽培結合精準農業 (Precision Agriculture, PA)的技術已行之有年。目前設施栽培之環境感測系統，大多採用以有限數量感測器的量測數據代表整個溫室內環境的資訊，此系統建構容易、成本低廉，但其精準度會隨著溫室面積擴大而下降，且無法精準呈現分佈不均之環境資訊，為解決此問題，可以利用無人飛行載具(Unmanned Aerial Vehicle, UAV) 結合環境感測器應用於農地環境資訊收集，無人機中四軸飛行器之易於操控與具有高度自由度的特性，恰可應用於量測溫度、溼度、照度等環境資訊以及植被指數等精準農業所需的相關的資訊。 在四軸飛行器的應用上，定位導航、路徑規劃與避障為三大重點，本研究將以無人機為載具搭載感測器，並在溫室內建立一室內定位系統。在飛行任務的路徑規劃上，由於溫室內環境設施為固定不變的障礙物，因此可透過決策系統以靜態路徑規劃的方式，來決定飛行機器人巡航的路徑，搭配決策控制系統進行感測路徑之規劃，達到巡航於溫室內的目的，並能避開溫室內的障礙物。在決定好任務路徑後，透過自動控制系統使飛行機器人能精準地依照規劃的路徑，完成自動巡航於溫室中，並收集環境資訊。 本研究建立溫室環境中可使用之室內定位系統，在定點定位實驗結果中，X軸平均誤差、Y軸誤差與E_a(絕對誤差)分別為0.09 m、0.10 m與0.14 m。在飛行自動控制的系統中，定點懸停定位的誤差表現上，X座標的均方根誤差RMSE (root-mean-square error) 為0.08 m，Y座標的RMSE為0.07 m，皆小於機身半徑(0.5m)，此定位系統可在室內環境下穩定追蹤靜態與動態目標。 在溫室的實地測試中，各錨點的距離資訊的檢量線相關係數皆達到0.99以上，在定點定位上，X軸、Y軸及E_a(絕對誤差)的平均誤差分別為0.12 m、0.08 m及0.15 m，與理想環境的實驗室條件下的結果相近，顯示在溫室中此室內定位系統亦可有著相當高的精準度，而在活動植床上進行的穩定度實驗， X座標與Y座標的RMSE分別為0.12 m、0.13 m，與靜態定點定位的表現相似，而Z座標的RMSE為0.05 m，顯示無人機在活動植床上可穩定飛行的能力，而在溫室實地飛行的測試中，無人機可成功沿規劃路徑飛行，並同時蒐集溫室的環境資訊，在資料庫中建立環境資訊的平面分布圖。 本研究以四軸飛行器為主體建立一溫室資訊偵測飛行機器人，以基於超寬頻技術建立室內定位系統，搭配決策控制系統進行感測路徑之規劃，達到巡航於溫室內的目的，並能避開溫室內的障礙物，將感測器之環境資訊與其位置資訊於網路資料庫中進行彙整與分析，以作為精準栽培之依據。

Precision agriculture technology has been applied to greenhouse production for some years. Greenhouse cultivation nowadays usually uses only one or a few numbers of sensors to represent the environment information for the entire greenhouse area because it is easy to install and low cost. To solve this problem, the applications of unmanned aerial vehicles (UAV) combined with environmental sensors to collect environmental information in agricultural field. Since the UAV is easy to control and has high degrees of freedom, it can be also applied to measure temperature, humidity, lighting information and vegetation indices for precision agriculture practice. In the application of unmanned aerial vehicles, position navigating, path planning and obstacle avoidance are the three major priorities. The objectives of this research are to build a flying robot for greenhouse environmental information measurements by the unmanned aerial vehicle, and to establish an indoor positioning system based on ultra-wideband technology. Since the obstacles in the greenhouse mostly are fixed structure, the decision-making system can determine the flight robot's cruise path by static path planning to achieve the purpose of cruising in the greenhouse, and to avoid obstacles in the greenhouse. After the cruise path is determined, the flying robot can accurately follow the path and collect environmental information simultaneously. The environmental information and location information will then be stored in a network database to integrate and analyze. Results of fixed-point positioning experiments show that, the average X-axis error, Y-axis error, and E_a(absolute error) are 0.09 m, 0.10 m, and 0.14 m. In the fixed-point flying test, the RMSE (root-mean-square error) of X coordinate is 0.08 m, RMSE of Y coordinate is 0.07 m. The positioning system built in this study can stably track static and dynamic targets in an indoor environment. In the field test in the greenhouse, the correlation coefficients of the calibration information of the distance information of each anchor point are all above 0.99. In the fixed-point positioning, the average errors of the X-axis, Y-axis, and E_aare 0.12 m, 0.08 m, and 0.15 m respectively, which are similar to the results under ideal laboratory conditions. It indicates that this indoor positioning system can also have a very high accuracy in the greenhouse. In the experiment of the fixed-point flying test on movable benches in greenhouse, the RMSE of X-axis and Y-axis are 0.12 m and 0.13 m, which are similar to the performance of static fixed-point positioning, while the RMSE of the Z coordinate is 0.05 m, showing the ability of the drone to stably fly on benches. In the field performance test in the greenhouse, the drone can fly along the planned path and collect the environmental information of the greenhouse at the same time, and establish a 2-D distribution maps of the environmental information in the database cloud system. In this research, the UAV was adopted as the main body to build a flying robot for greenhouse environmental information detection. Establishing an indoor positioning system was based on ultra-wideband technology. The decision-making control system can be used to plan the cruise path to achieve the purpose of cruising in the greenhouse and to avoid obstacles in the greenhouse. The environmental information and location information of the sensors are then integrated and analyzed in a network database to provide precision cultivation suggestions in the greenhouse.