Research And Implementation Of Intelligent Learning Control For Two-wheeled Robot

Two-wheeled robot,whose control principle of attitude balance is similar with human,this kind of robot that has the characteristics of multi-variable,nonlinear,strong coupling and parameter uncertainty is an unstable system.So,how to make the twowheeled robot maintain the balance of the human body as a person under all kinds of outside interference external interference is crucial.In the same while,the high intelligence and good human-computer interaction mode,which can make the two –wheeled robot has high application value and wide application prospect with the rapid development of science and technology today.In this thesis,two-wheeled robot model is constructed by using mechanism modeling and system identification respectively based on the two-wheeled robot physical system of the Android smart phone;and then,the intelligent learning control is researched followed by self-learning control algorithm and self-learning control algorithm based on LQR is designed.Finally,human-computer interaction system based on mobile phone is designed.The main contributions can be organized in the follows:First: Mechanism modeling and system identificationThe physical system two-wheeled robot that can be given motion commands by intelligent mobile phone is constructed,mobile phone can communicates with robot via Bluetooth module.For getting the appropriate control method,mathematical model of the robot is constructed.First,the physical system of two-wheeled robot are simplified and assumed,dynamic model of two-wheeled robot is built based on Lagrange energy equation method.Second,the zero input response and zero state response simulation experiments are designed to verify the correctness of the model.Third,controllability and observability of the model are analyzed by linearization at the equilibrium point,and the PID classical control algorithm is used to simulate the model,and the correctness of the model is verified again.In order to get a more accurate model,the recursive least squares method is used to identify the system and estimate parameters by collecting the actual input and output data of two-wheeled robot.Second: Self-learning controlFor tackling the issue of the disadvantages of traditional control algorithm,selflearning control algorithm is proposed,this kind of algorithm has the advantages of simple structure,none accurate mathematical model,none complex rules,it can train the optimal control parameters in the current environment through continuous learning and self-regulation,and get the best control effect of two-wheeled robot.In the simulation and physical experiments,impulse and white noise interference and changing of the weight of body is applied to test the performance of robust,LQR is taken to make comparative trial to verify the effectiveness of the self-learning control algorithm.Third: Self learning control based on LQRFor improving the performance of control,an algorithm of self-learning control based on LQR is proposed.The algorithm initializes the initial value of self-learning algorithm by LQR optimal parameters can improve self-learning algorithm control effect and greatly reduce the amount of calculation of self-learning algorithm,it can learn and train a few times to get better control effect.In the simulation and physical experiments,by applying impulse interference and white noise interference and change the weight of the body,and in contrast the self-learning algorithm,the self-learning control algorithm based on LQR has better performance compare with using selflearning algorithm or LQR only.Forth: The application of remote control based on Android SmartphoneFor dealing with the disadvantages of the traditional infrared remote controller,a remote control application of two-wheeled robot,which make the control of twowheeled robot diversified and intellectualization by applying all kinds of functions and sensors of Android smart phones based on Android smartphone is proposed.The twowheeled robot motion direction and the state can be controlled by touching the interactive keys,calling gravity sensor and speech recognition,meanwhile the smartphone can receive the specified data that upload on the robot,such as the pitch angle and speed,controller can also set parameters of control procedures,for example the running speed of the robot and the parameters of control algorithm;the environment around the robot can be taken photos and record in its memory,and the robot can display the position of the robot in the program built-in electronic map and make path planning when installing the program's Android smartphone holder on the robot.