Research On Indoor Navigation Of Quadrotor Aircraft Based On Monocular Vision

Due to the advantages of vertical takeoff and landing and simple structure,quadrotor aircraft has become a hot research field in recent years.The current quadrotor aircraft navigation is mainly using the Global Position System(GPS).However,GPS signals may be disturbed or even lost in buildings,forests,and other sheltered environments.This will make it impossible for the navigation of the quadcopter to receive navigation information or receive incorrect navigation information,resulting in unnecessary life and property loss.At present,thanks to the development of visual sensors,the accuracy of optical navigation systems are getting higher and higher.The volume is more convenient to install and carry,and the cost is relatively low.It is considered to be the ideal solution to solve the problem of autonomous navigation of quadrotor aircraft in the absence of GPS environment.In this thesis,based on the optical flow navigation algorithm,how to use optical flow to navigate the quadrotor aircraft in the indoor without GPS environment is studied.Firstly,this thesis describes the optical flow algorithm and the development of the quadrotor aircraft in detail.Then the quadrotor aircraft was modeled.Based on the establishment of a quadrotor aircraft dynamics model,a quadrotor aircraft co-simulation platform was first established using the dynamics software ADAMS and the mathematics software MATLAB/Simulink.Then,a multi-loop compensation Proportion Integration Differentiation(PID)controller was designed and simulated using MATLAB.The feasibility and effectiveness of the control algorithm were verified.And using the complementary filter fusion optical flow data and inertial unit data to obtain accurate positioning information,using the Pixhawk flight control and PX4 FLOW monocular camera to build a quadrotor aircraft platform for hover flight test.Through actual flight tests,it can be seen that the multi-loop PID algorithm can achieve fixed-point hovering tasks in conjunction with data fusion.For the problem of the optical flow algorithm feedbacked from the experimental results,an optical flow algorithm based on Oriented FAST and Rotated BRIEF(ORB)was combined with Random Sample Consensus(RANSAC)for improvement.The improved algorithm is compared with simulation experiments.The experimental results show that the algorithm can accurately estimate the position.