Research On Robust Control Technology Of Quadrotor Unmanned Aerial Vehicle Under Urban Environment

The quadrotor unmanned aerial vehicle(QUAV)which is capable of vertical take-off and landing,hovering and flying in low speed can be categorized to the unmanned rotorcraft family.Compared to the conventional unmanned helicopter,the quadrotor has lower cost and easier to fix.On this characteristics and advantages,the quadrotor has been widely used in military and civilian areas in recent years.Chinese companies have done a lot of rearch for QUAV,which is helpful for the QUAV consumer market,at the same time,these companies have take more and more important role in UAV industry.This paper focuses on the flight control,sensors and its pre-processing,double eyes system,obstacle avoidance and obstacle detection,which is aimed at robust control for flight.The main research contents are listed as follows:Firstly,the principle for flight of QUAV is analyzed in detail,and the coordinate systems of both world and QUAV are established,the transfer matrix between these two coordinate systems is calculated at the same time.The nonlinear dynamics are established based on the transfer matrix,and for further facilitate the engineering realization and controller design,the linear model expressed by transfer function is obtained using LPV linearization technique.Secondly,considering of model uncertainties and external unknown disturbance on QUAV,a robust control method based on H infinite loop shaping is designed.Left coprime factorization is adopted due to the model uncertainties.Based on the analysis of solving methods of robust stability problem,the concept and design procedure of loop shaping methods are introduced,and the design of compensator is further discussed.In terms of practical use,the order of the robust controller is reduced,and the results of simulation show that the designed robust controller is effective.Thirdly,in order to get adequate information of urban environment,the fisheye camera is considered as the main sensor with the analisis of model and characteristics of fishey camera.Two methods of image correction are compared under experiment,and a new kind of image correction method is proposed to be superior,which is based on three-dimensional model.At the same time,a new kind of fisheye camera calibration based on genetic algorithm is proposed.With the compare between the traditional method and the new kind of method,the simulation shows the effectiveness of the new method of calibration.Following,a kind of obstacle recognize module is designed,which is divided into two parts,they are foreground detection module and obstacle classify.The vibe algorithm is used in the foreground detection,and the experiment result shows its validity.To realize obstacle classification,a kind of classifier based on SVM algorithm is designed.The SURF algorithm is used to get the feature points,and K-means algorithm is used to cluster feature points,in order to get the feature vector of image,the BOW mode is applied.In the simulation,several kinds of kernel functions are compared,the result shows that the classifiler with the polynomial kernel has the best recognization accuracy.Finally,obstacle avoidance module is designed under urban environment.There are global path planning and local obstacle avoidance in this system,and A-star algorithm is applied into the global path.Due to the disadvantage of local minimum point,the virtual obstacle is proposed to optimize the artificial potential field algorithm as the local obstacle avoidance method.To get the information of distance,double eye system is designed.In the double eye system,ORB algorithm is used to extract feature,and PROSAC alrorithm is used to optimize feature matching with the comparation of RANSAC alrorithm.The simulation shows that the designed double eye system has good performance in distance detection.