Experimental implementation of fuzzy vision- based tracking control of quad-rotor Hemjyoti Das 1 , Aniket Samrat Mazumdar 1 , Rajeeb Dey 1 , Lintu Roy 2 Department of Electrical Engineerings 1 , National Institute of Technology, Silchar Department of Mechanical Engineering 2 , National Institute of Technology, Silchar Email:- hemjyoti.nit@gmail.com ,rajeeb_de@ieee.org Abstract- This paper presents a novel approach to detect and track an object using a quadrotor-UAV. The proposed system mainly consists of two parts-(i) Object detection and tracking using histogram backprojection and CAMSHIFT tracker, (ii) Fuzzy Proportional and Fuzzy Proportional-Derivative controller for controlling the drone. We implemented our algorithm using ROS (Robot Operating System), OPENCV library and MATLAB programming environment. Keywords— CAMSHIFT, ROS, Fuzzy Logic, Fuzzy-PD Controller, MATLAB, Open-CV I. INTRODUCTION In the recent years, there has been a lot of ongoing research in the field of Aerial Robotics. The quad-rotor which is a four rotor micro-aerial vehicle has got immense popularity among various research groups due to its ease of handling compared to the fixed wing vehicles. Various algorithms related to path planning, machine learning, computer vision , control systems , etc are now being implemented on drones to increase the wide range of applications of drones. A wide variety of research is being done by various universities and research groups on the vision based control of drones like, in [1] a camera based navigation system for the Parrot AR Drone have been designed using visual SLAM, Extended Kalman Filer and PID Controller. In [2], an image based visual servoing (IBVS) controller using OpenTLD tracker and PD controller have been developed. Quadrotors are also used nowadays for tracking of Marine Animals as, in [4] Hue and Saturation values for calculating the Histograms for object identification and tracking have been used. In [5], Hough transform was used for detecting players in an indoor soccer game using quadrotor. Similarly in [3], hough transform is used for autonomous object tracking. In [7], a new control approach is designed using switching control which utilizes Edge Detection and Optical Flow algorithm. In [8], a quad-rotor UAV is simulated for tracking a moving ground object in the presence of noise. In [9], a new algorithm called dagger algorithm is proposed for high speed autonomous flight through dense forest environment. Nowadays landmarks have become very effective tool for localization as discussed in [10]. Fuzzy controllers are also being used in different systems like in [11], they have implemented Fuzzy-PD+I controller for controlling a Magnetic Ball Levitation System .In [15], they successfully implemented a fuzzy controller for target tracking on a Pelican quad-rotor running on an Intel Atom Board PC. The main aim of this experiment is to autonomously detect an object and track it using Fuzzy based Controller. The paper is organized as follows: Section II describes the hardware platform and the software architecture used in the experiment, Section III discusses the object detection and tracking algorithm. In section IV, controller design for the Parrot AR Drone is discussed. Finally the experimental results are shown in section V. II.EXPERIMENTAL PLATFORM For the realization of this project, the Parrot AR drone quad- rotor is used. The AR-Drone 2.0 [14] is an electrically powered quad-rotor intended for augmented reality games. It is preferred to other drones due to its cheap price, robustness of hardware, ability to execute basic control commands through its on-board softwares and compatibility with Robot Operating System (ROS). The AR Drone's SDK functions can be translated easily into ROS nodes, which allows us to integrate and compare our project with existing research. Fig 1: Parrot AR Drone 2.0 and its coordinate system A. Hardware Description [14] 1) Support structure and body: The body and support structure of the Parrot AR Drone 2.0 quad-rotor are made out of plastic and carbon fiber respectively because of their lightweight and resistance to mechanical wear out. This keeps the quad-rotor's weight to approximately 380g or 420g, depending on the hull of the quad-rotor.