基於影像系統與慣性測量單元之標記導航四旋翼機的視覺伺服最佳控制

Abstract

近年來，有關微型飛行器（Micro Aerial Vehicle）的研究日益增加。四旋翼機是一種擁有四個旋翼的飛行器。它能夠提供穩定的特技飛行，並且能航行至地面機器人或人類無法到達的地方以執行一些高難度以及危險任務或工作，譬如農場監測和地雷檢測等等。

本研究的目標是開發出一套控制系統以讓Parrot AR.Drone 2.0四旋翼機進行各種視覺伺服任務，例如基於地面標記以及置於底部Raspberry Pi攝影機模組的影像回饋之起飛、停懸、追蹤以及降落等動作。這種可以執行上述任務的自主飛行系統將可以為後續的研究帶來更多可能性，例如利用裝在飛行器上的機械手臂以抓取地面上的物體，以無線方式給地面上的傳感器充電，無需人工干預地自動降落於停靠站或充電站以延長其飛行時間以及進行空中監視等等。該自主飛行系統採用一張信用卡大小的單板電腦Raspberry Pi 2並搭配Robot Operating System（ROS）作業系統來實現本論文所提出的各種演算法。本論文提出了LQG（Linear-Quadratic-Gaussian）最佳控制法作為飛行器的橫軸（lateral axis) 與縱軸（longitudinal axis）速率的控制且使用PD（Proportional-Derivative）控制法作為飛行器的橫軸、縱軸與垂直軸（vertical axis）位置以及偏擺（yaw）方向的控制。在設計控制器前，本論文先對AR.Drone 2.0進行系統辨識（System Identification），然後使用MATLAB軟體來對系統進行分析與設計控制器,最後再使用Simulink模擬設計好的控制器。本論文描述了四旋翼機控制器的設計以及基於開源電腦視覺（OpenCV）的標記（Visual Marker）偵測方法，並且把Raspberry Pi 2搭載於AR.Drone四旋翼飛行器上，以進行各種視覺伺服任務的測試。

In recent years, there is an increased interest on the research of Micro Aerial Vehicles (MAVs). Quadcopter is a class of four-rotored aerial vehicles which has the capabilities to provide stable acrobatic flight, navigate into places that ground robots or human cannot reach and perform tasks which are difficult or dangerous for human being such as agricultural surveillance and explosive landmines detection. The goal of this thesis is to present control systems for the AR.Drone 2.0 quadcopter by Parrot to perform visual servoing tasks such as take-off, hovering, following and landing based on the vision feedback from Raspberry Pi camera module attached to its bottom and a visual marker placed on the ground. By enabling an autonomous MAV to perform the aforementioned tasks, it will open a lot of posibilities such as following or picking up ground objects, charging wireless ground sensors, landing on docking or charging station without human intervention to increase its flight time, aerial monitoring and inspection. All the algorithms of this system is implemented on a single-board computer, Raspberry Pi 2 which runs the Robot Operating System (ROS). This thesis has proposed the LQG (Linear-Quadratic-Gaussian) controller for the velocity control of quadcopter’s lateral and longitudinal movement while using PD (Proportional-Derivative) controller for the position control of quadcopter’s lateral, longitudinal, vertical movement and also its yaw rotation control. This thesis has used System Identification method to obtain the system model of AR.Drone 2.0. Then MATLAB software is utilized to analyze and design the controllers and finally the designed controllers are simulated in Simulink. This thesis describes the development, testing and implementation of a system which uses Raspberry Pi 2 as onboard computer to a real quadcopter, AR.Drone 2.0. The system can perform visual marker detection by using Open Source Computer Vision (OpenCV) library for image processing, and also perform the automated visual servoing tasks.