Dynamic Modeling And Research On Adaptive Control Strategy Of A Front-wheel Drive Bicycle Robot

The front-wheel drive bicycle robot is an under-actuated non-integrated system with lateral instability, the two wheels are longitudinal and has non-sliding contact with the ground, meanwhile the dynamics modeling possesses a kind of symmetrical feature, and the dynamic characteristics are complicated. Therefore, it is difficult to control the bicycle robot stable, and the dynamic modeling and self-balancing control should be the hot topic in robotics for a long time.This dissertation discusses a front-wheel drive bicycle robot with no regulator weight. The research work about this robot concentrates on the dynamics, the prototype hardware construction, the 90° handle-bar track-stand motion and the stable-balanced rectilinear motion as well. The 90° handle-bar track-stand motion means that the bicycle robot can maintain the balance itself when the speed of the bicycle is zero and the handle-bar turning angle is 90°.The prototype experiments on vary balance motions are intentionally highlighted in this dissertation. The detail work of the dissertation is as follow:At first, based on Lagrange Equation, this dissertation build a general dynamic model of arbitrary front-bar turning angle by analyzing the instant rotational axis and the turning radius of the bicycle robot. According to the proposed dynamic model of arbitrary front-bar turning angle for the bicycle robot, a dynamical model of the stable-balanced rectilinear motion and a dynamical model of 90° handle-bar track-stand motion were built. Two dynamical models were linearized based on Taylor series expansion. And the observability, controllability, stability and controllable angle of the linearization models are analyzed.Secondly, two LQR controllers were designed for the stable-balanced rectilinear motion and the 90° handle-bar track-stand motion of the bicycle robot. The results of the Matlab numerical simulation results showed that the LQR controller has a good effect on the dip of the bicycle robot. The feedback gain matrix of the LQR controller in the linear model is applied to the nonlinear model for numerical simulation. By comparison between the simulation results, we found that the feedback gain matrix of the LQR controller under the linear model could be used in nonlinear systems,and the control effect went worse. The control range of bicycle robot is reduced and the times using for a stable system is prolonged. The changes mainly caused by giving away some the nonlinear term from the linearization procedure of system model.Thirdly, two adaptive fuzzy controllers were designed for the stable-balanced rectilinear motion and the 90° handle-bar track-stand motion of the bicycle robot. The Matlab numerical simulation results showed the efficacy of the adaptive fuzzy control strategy. From the simulation results, it can be seen that the control effect of adaptive fuzzy controller is better than LQR controller, parameter adaptive ability and anti-interference ability are also stronger.Fourthly, two RBF neural network adaptive controllers were designed for the stable-balanced rectilinear motion and the 90° handle-bar track-stand motion of the bicycle robot. The Matlab numerical simulation results showed the efficacy of the RBF neural network adaptive control strategy. From the simulation results, it can be seen that the control effect of RBF neural network adaptive controller is better than adaptive fuzzy controller, parameter adaptive ability and anti-interference ability are also stronger.Finally, a physical prototype of the front-wheel drive bicycle robot was constructed with two driving DC motor to provide driving force and gear reducers to transfer motion, and the embedded controlling system hardware of the prototype was emphasized. The controlling system utilizes digital signal processor (DSP) TMS320F28335 as the core controller and the datum acquired module, micro controller chip (MCU) C8051 F020 as motors controller, and adopts inertial measurement unit (IMU), encoders, Hall current sensors and supper sonic sensors to be sensors modular, and uses CAN, SPI and RS232 bus to exchange datum between processors. Based on this prototype, an adaptive fuzzy controller of the stable-balanced rectilinear motion of the bicycle robot is proposed. The experimental results demonstrate the stability of the prototype and the actual control effect of the adaptive fuzzy controller.