Autonomous Localization And Nonlinear Control Of Rotor Drone

In recent years,the rotor drone has become a major research area in the field of drones with the flexibility of maneuverability and the capability of vertical taking-off and landing,fixed-point hovering and other advantages.It puts high requirements for performing tasks of the rotor drone in the military or civilian field.To verify different localization algorithms and autonomous flight control strategies,it will be convenient to use small size quadrotor unmanned aerial vehicle(UAV)to build the flying testbed.It is very difficult to obtain a very accurate dynamic model for the rotor drone system due to its strong nonlinear properties,state coupling,and static instability.To ensure the safety and stability of the UAV flight,it brings very high requirements for the control design of the rotor drone system.At present,the research on the control problem of the quadrotor UAV has been widely investigated,but less research work has been performed on the unmanned helicopter under unknown external disturbances.Therefore,this thesis mainly works on the localization of the quadrotor based on small-size on-board Lidar,and reinforcement learning based nonlinear control design for a small-size unmanned helicopter.The main results of this thesis are listed as follows:1)This thesis presents a Lidar Simultaneous Localization and Mapping(SLAM)algorithm based on the graph optimization theory for the quadrotor UAV,which is employed to deal with the inaccuracy of the two-dimensional Lidar SLAM and the long time-consuming of the traditional Lidar SLAM algorithm due to attitude variation of the quadrotor UAV during the flight.The proposed algorithm firstly combines the Lidar data with the IMU data,and then optimizes the accurate pose estimation and construct the local submap by using the least squares optimization and a motion filter.At the same time,the Branch-and-Bound(BB)algorithm is developed for the closed-loop back-end optimization,which speeds up the efficiency of searching for the optimal solution.Finally,the accuracy and real-time performance of the localization algorithm are verified by real-time experiments.2)Based on the proposed Lidar SLAM algorithm,a flight testbed of a quadrotor UAV system based on a Hokuyo laser range finder is developed.The system can realize hovering flight,trajectory tracking and semi-autonomous flight control in the indoor environment and can monitor the drone map construction in real-time in the ground station through Wi Fi data communication.The experimental results show that the localization control accuracy of the UAV in the indoor environment reaches 0.1m,and can obtain an accurate map of the environment.3)This thesis presents a new nonlinear control law based on the combination of reinforcement learning(RL)and Super Twisting methodology for the attitude control of a small-size unmanned helicopter,which is subjected to modeling uncertainties and unknown external disturbances.The proposed control law only uses input and output data of the helicopter to train the Actor-Critic(AC)neural networks to compensate for modeling uncertainties.Then a nonlinear robust controller based on Super Twisting methodology is developed to compensate for the unknown external disturbances.The Lyapunov based stability analysis is used to prove that the attitude error of unmanned helicopter can converge to zero in finite time.Finally,the proposed control law is verified on a self-built hardware in the loop testbed.The experimental results show that the proposed control law can achieve good control performance together with good robustness to modeling uncertainties and wind disturbances.