A Study On Some Issues For Bicycle Robot Nonlinear System

The bicycle robot is a natural unstable nonlinear system. It has the characters of static instability and dynamic stability. And the inherent nonlinear and deeply coupling characters of bicycle robot make the control of it much more difficult. The bicycle robot dynamic modeling and control is a difficult problem in robotics. This kind of bicycle robot is the first one presented in our country. One hand, there are a lot of the means of bicycle robot dynamic modeling. At present, the bicycle robot dynamic models presented in world are not consummate. So it has high value to present some accurate and general dynamic models for bicycle robot. On the other hand, the control algorithm is the one of the key problem to control the bicycle stably at the view of control theory. Much of the overseas research is on the phase of computer simulation or design controller for bicycle robot based on linear control theory. And much more factors should be considered when we have an experiment of control a real bicycle robot. In the paper, three kind of bicycle robot dynamic model are presented. After analyzing those affine nonlinear systems, their corresponding control controllers are designed. And two kind of real bicycle robot were designed. The control experiments were performed based on the real robot. The main contents can be summarized as follows.1. A kind of steering control single-input single-output dynamic model for bicycle robot is presented based on Lagrange method. A fake linear system was presented based on the approximate linearization method. And the controller was designed for the linear system. The nonlinear dynamic model was expressed based on the format of SISO affine nonlinear system. Then the system was represented into the standard Byres-Isidori format base on the diffeomorphism transform. The standard format system has a non-minimum-phase zero dynamic subsystem. And the system was assuaged base on the central manifold theorem. A fuzzy sliding-mode controller was designed for the system for better performance.2. Taking the steering torque and the driving torque of rear wheel as system inputs, a kind of dynamic model was presented based on Lagrange method. And the controller was designed for the system to achieve the goal of decoupling and linearization based on input/output linearization.3. Similarly, the steering torque, the driving torque and the mass balancing torque were taken as system inputs, a kind of dynamic model which has 3 inputs and 3 outputs was presented based on Lagrange method. And the control algorithm was designed for the system to achieve the goal of decoupling and linearization based on input/output linearization.4. A new kind of dynamic model was presented based on Lagrange theorem. The steering angle, tilting angle and the pitching angle are taken as the inputs of the model. And the control algorithm was designed for the system to achieve the goal of decoupling and linearization based on input/output linearization.5. Two new types of tilting angle sensor were designed to measure the tilting angle accurately and sensitively. One was designed based on MEMS rate gyroscope. And the other was designed on the change of the pulse frequency aroused by the change of inductance. And a kind of sampling system was designed to achieve the goal of sampling the robot gesture from sensors when we have outdoor experiments.6. The real bicycle robot was designed. And the digital control platform was built of the robot. Some experiments were performed base on the robot. Take the bicycle robot MIMO affine nonlinear system as example, the experiments data were analyzed and contrasted with the simulation results.