Two-wheel Self-balancing Robot Modeling And Intelligent Control

Two-wheel self-balancing robot is one of the most important imitate-life systems among wheel-pattern robots, it has many advancement, its structure is simple, the volume and weight of it is small, as well as the movement is smart, so its foreground is ocean in the application of society and industry, as vehicle, the developing room of it is very wide, It's of great commercial and learning value to research it, In allusion to the absolute instability of two-wheel self-balancing robot, this paper researching on the balancing control problem of the system. basing on the analyzing and inspecting two-wheel self-balancing robots home and abroad, this paper introduce the hardware structure and the primary components of the object in detail, analyzing the principles of the electrical system, software system and state measuring system. Modeling the system by Newton mechanics analysis and Lagrange equation, and analyzing the stabilization, controllability, observability and relative controllability of the system qualitatively. Coming to the control methods, firstly, In allusion to the linear model of the two-wheel self-balancing robot, using typical modern control theory pole place to design state feedback controller, analyzing the initial obliquity and displacement control by simulation, secondly, using least square method to identify linear controller parameters of state feedback, that is multi-linearizing the nonlinear model first, then gaining the I/O data through the model collection by pole place method, using this data, through least square method to identify the parameters of the state feedback matrix, then comparing the state feedback control identified by least square method and the pole place control. Lastly, basing on the linear control, this paper design intelligent controllers ulteriorly: fuzzy-PD controller designing, fuzzy-PD control which is constituted by displacement fuzzy control and obliquity PD control, giving a method to estimate I/O proportional factors, and analyze the impact of control when the factors change; fuzzy neural controller designing, adopting adaptive neuron fuzzy control to realize sample data control, using adaptive neuron fuzzy inference system to train membership function to gain fuzzy control rules, the training data come from I/O data of pole place linear controller, using the fuzzy neural network as three free degree model controller, and simulating the control effect of it.