Adaptive Learning Control Of A 2-DOF Helicopte

As a modern and very characteristic aircraft,unmanned helicopters,with the advantages of vertical take-off and landing,hovering and excellent mobility,have extensive applications in material transportation,military reconnaissance,rescue and disaster relief.However,the helicopter is a highly nonlinear and multiparameter complex coupled system with system uncertainties and parameters uncertainties.In addition,input nonlinearity and output constraints brought about by the physical-mechanical structure,as well as external disturbances from unknown environment,are factors that can cause adverse responses or even system instability to the helicopter system.With the development of artificial intelligence,to achieve intelligent and high-precision control of unmanned helicopters has become an important research area,which has attracted the attention of many scholars in recent years.Adaptive neural network control is a common intelligent control method.However,considering that the application scenarios of unmanned helicopters are often complex and variable and the execution tasks are usually regular,the traditional adaptive neural network control in dynamic environment will have some issues,such as insufficient adjustment ability,expensive computational waste and long convergence time.Hence,in this paper,based on adaptive learning control strategy,the following aspects will be investigated for the high performance control issue of a 2-DOF helicopter nonlinear system.1、An adaptive reinforcement learning control technique is studied for the 2-DOF helicopter nonlinear system.First,the evaluation and execution mechanisms are built according to the framework of reinforcement learning.The evaluation mechanism can give an appropriate evaluation value based on the set cost function and the current system states.The execution mechanism can integrate the evaluation value and the approximation error of system uncertainty to make the corresponding control strategy.By introducing a more comprehensive evaluation mechanism,the adaptive law of the neural network can be improved,and then the real-time adjustment capability of the controller can be effectively improved.2、An adaptive broad learning control for the 2-DOF helicopter nonlinear system is investigated.First,the traditional radial basis neural network is improved based on the method of broad learning,so that the neural network can selectively add hidden nodes by judging the relationship between the current input vectors and the hidden layer nodes.This is to ensure that the input vector can be included in the tight set,and improve the generalization capability of the neural network,such that the system multi-source uncertainties were better handled.3、Deterministic learning control technique for a 2-DOF helicopter nonlinear system is investigated.Considering that helicopter usually perform the same or similar tasks,the repetitive adaptive iterative process causes a huge waste of computational cost.Meanwhile,the influence of the input gap on the tracking performance is also considered.This paper implements the identification of the nonlinear system model of helicopter and the parameter recognition of the inverse model of the backlash based on the deterministic learning theory,and stores and utilizes the learned knowledge to construct a learning controller with low computational cost and high performance.