The two self-balancing robot is an important member of wheeled mobile robots. It has the advantages of small size, light weight and flexibility and is valuable in practice. With its nonlinear, multi-variable, strong coupling features, self-balancing robot also has a high theoretical significance. This paper analyzes the robot’s balance control problem on the basis of labrotary’s prototype robot and the main work is as follows: First, the overall structure of the self-balancing robot is designed. The overall design contains the mechanical structure design, electrical structure design and software design. The processor unit, sensor unit and implementation mechanism unit is introduced. Second, the pose measurement system is designed. Pose measurement system plays an important role of self-balancing robot. The optical encoder measures the robot’s position and speed, the MEMS accelerometers and gyroscopes measures the inclination and angular velocity respectively. In order to overcome the inherent shortcomings of inertial sensing elements, Kalman filtering is used to fuse accelerometer and gyroscope information and more accurate pose information is acquired. Third, the self-balancing robot system is modeled. On the basis of simplification assumptions, by measuring physical and geometric parameters of self-balancing robot, the nonlinear system is modeled using Lagrangian dynamics equations. Then the model is linearized and decoupled to two single-input linear systems, so the difficulty of designing a controller is reduced. Fourth, two preliminary controllers are designed. PID is the most commonly used control method. This paper firstly designs a PID controller with adjusting PID parameters by experience, and simulation results show the PID controller is not appropriate for a multivariable systems. An LQR controller is designed using Linear Quadratic Regulator (LQR) method, the simulation comparison experiments and physical experiments on the real system both prove that the LQR is effective. Finally, the fuzzy controller and neural network controller is designed to increase the system’s robustness and adaptive capacity. The LQR is essentially linear feedback controller and is not fit for a complex nonlinear system, especially when the system is subject to outside interference or a sudden change of the system parameters. This paper designs the fuzzy LQR controller and neural network LQR controller on the basis of LQR method. Simulation experiments show th...
修改评论