Study On Neural Network Control Of Two-DOF Helicopter With Non-smooth Nonlinearit

In recent years,small unmanned helicopters have been widely used in military,commercial and agricultural fields due to their outstanding maneuverability and portability,and the rate of personal ownership has also been increasing.However,due to the mechanical design and unique structure of small unmanned helicopters,they exhibit highly nonlinear and strongly coupled characteristics,and there are also uncertainties in the system.Therefore,designing a reasonable controller to ensure its stability is very challenging.Considering the unique mechanical structure of small unmanned helicopters,there are non-smooth nonlinear characteristics such as input dead zone,input saturation and hysteresis nonlinearity in the system.These non-smooth nonlinear characteristics will constrain the control performance of the system,resulting in the inability of the helicopter system to achieve precise control.In addition,during long-term operation of the small unmanned helicopter system,wear and tear on the system’s actuator components may cause failures.If not dealt with in a timely manner,it may reduce the stability of the system or even cause safety accidents.Furthermore,when using a computer to remotely operate a small unmanned helicopter,control signals are transmitted between nodes such as the system’s controller and actuator via wireless communication.However,due to the limited bandwidth of the communication channel,communication data often experiences delays and packet loss during transmission,which not only wastes communication resources but also affects the control performance of the system.In summary,it is necessary and meaningful to design a reasonable controller to solve the problems existing in small unmanned helicopter systems.This paper uses a two-degree-of-freedom helicopter platform as the research object,which has typical characteristics of small unmanned helicopters and is very convenient for data collection.It is the best experimental platform for verifying the neural network(NN)control algorithm of small unmanned helicopter systems.On this platform,the NN control of a twodegree-of-freedom helicopter system with non-smooth nonlinearity is mainly verified,and a reasonable control strategy is designed in combination with adaptive control.The main contents of this paper are as follows.1.For a two-degree-of-freedom helicopter system with input dead zone and actuator failure,a NN control algorithm is proposed in combination with adaptive control.Firstly,a radial basis function neural network(RBFNN)is used to estimate the uncertainties in the system.Secondly,bounded estimation,smooth functions and adaptive auxiliary parameters are used to solve the combined effects of actuator failure and input dead zone.Then,Lyapunov stability theory is used to ensure that all signals in the system are bounded.Finally,the effectiveness and reliability of the controller design are demonstrated through simulation and experimentation.2.For a two-degree-of-freedom helicopter system with input saturation,in order to reduce the communication burden of the system’s input signals and avoid wasting communication resources,a NN quantization control strategy is developed in combination with input quantization.Firstly,a hysteresis quantizer is used to eliminate the jitter phenomenon caused by the quantization of signals.Secondly,a RBFNN is used to estimate the uncertainties in the system,and bounded estimation,smooth functions and adaptive parameters are used to solve the effects of saturation nonlinearity and quantization nonlinearity.Then,Lyapunov stability is used to prove that the system signals are bounded.Finally,through simulation and experimental analysis,it is verified that the control strategy has good control performance.3.For a two-degree-of-freedom helicopter system with hysteresis nonlinearity,a NN eventtriggered control is designed in combination with an event-triggering mechanism.Firstly,an event-triggering mechanism is introduced to alleviate the communication burden of system signals during transmission.Then,a RBFNN is used to approximate the uncertainties in the system.In addition,smooth functions and bounded estimation are used to solve the combined effects of hysteresis nonlinearity and network measurement errors in event triggering.Lyapunov stability theory is used to prove that all signals in the closed-loop system are bounded.Finally,the superior performance of the designed control algorithm is verified through simulation and experimentation.