Design Of The Control System For The Quadrotor Manipulator And Study On The Aerial Grasping Application

Unmanned aerial vehicles(UAVs),also known as drones,are one kind of robots.Thanks to such autonomous aircrafts,the application of robots has been extended to 3D space.Quadrotor helicopters are one of the most popular UAVs.They have been successfully used for reconnaissance,navigation,aerial photography,etc.Nowadays,it has been suggested that they can be used for rescue,express delivery,serving in restaurants and so on.These applications involve direct contacts with circumstances.Such interaction can be realized by mounting a robotic arm under the quadrotor.However,this will bring about more nonlinearity and states' coupling to the essentially nonlinear and under-actuated quadrotor system.Besides,manipulation on the surrounding objects leads to uncertainties so that the modelling could be unprecise.With more difficulties in controller design,more stability,rapidity and accuracy need to be guaranteed for its real-life applications.This dissertation focuses on one typical manipulation: aerial grasping.Different kinds of control methods are studied and presented.A large number of simulations have been done to compare and evaluate these control algorithms.The aim is to control the system so that it can follow the desired trajectory as precise as possible while grasp an object during flying.Firstly,the full-state model of the quadrotor with a robotic arm is established by Euler-Lagrangian method,followed by a simplification of it.Then,a classical PID controller is designed based on the simplified model for trajectory following.Afterwards,2 model predictive control algorithms are studied,that is,model predictive control based on state space representation and adaptive generalized predictive control with input-output linearization.Step responses of these 3 controllers are obtained and the performance is compared.Finally,an aggressive trajectory for aerial grasping is planned by Minimum Snap method and these 3 controllers are all applied for tracking the desired path with the quadrotor performing grasping.Simulation results show PID controller is not ideal for nonlinear systems with coupling states while the plant with nonlinear model predictive control is able to track reference states with reasonable control inputs value.What's more,compared with the model predictive control based on state space representation,the adaptive generalized predictive control with input-output linearization is a controller with adaptiveness and robustness and it requires less precise model.Even with abrupt variation of the system parameters,the quadrotor with a robotic arm can still follow the desired path.This is a better choice for aerial grasping manipulation.The control method and simulation results presented in this paper provide theoretical basis for UAV-manipulator system design and experiments.It can also be extended to other research fields such as space manipulators,suspended-load UAVs,etc.It has great significance in deploying UAVs to do more real work and bring more convenience to our daily life.