Test platform design and control of a bicycle-type two-wheeled autonomous vehicle
Bicycle dynamics and behaviors have been vastly studied through modeling and simulation. Due to the complexity, software models are often assumed subjecting to di erent nonholonomic constraints in order to simplify the models and control algorithms. A real life autonomous bicycle faces perturbances from the road, wind, tire deformation, slipping among other external forces. Limitations of simulations will not always allow these to apply. All these issues make the autonomous bicycle research very challenging. To study the bicycle control problems a few research results from the literature are reviewed. A nonlinear bicycle model was used to conduct control simulations. Model based nonlinear controllers were applied to simulate the balance and path tracking control. A PID controller is more practical to replace the non-linear controller for the balance control. Simulation results of the di erent controllers are compared in order to decide the proper control strategies on the hardware platform. The controller design of the platform complies with practicality based on the hardware con guration. Two control schemes are implemented on the test platform; both are developed with PID algorithms. The rst scheme is a single PID control loop in which the controller takes the roll angle feedback and balances the running platform by means of steering. If the desired roll angle is zero the controller will try to hold the platform at the upright position. If the desired roll angle is non-zero the platform will be balanced at an equilibrium roll angle. A xed roll angle will lead to a xed steering angle as the result of balance control. The second scheme is directional control with balance consisting of two cascaded PID loops. Steering is the only means to control balance and direction. To do so the desired roll angle must be controlled to achieve the desired steering angle. The platform tilts to the desired side and steering follows to the same side of the tilt; the platform can then be lifted up by the centrifugal force and eventually balanced at an equilibrium roll angle. The direction can be controlled using a controlled roll angle. Many implementation issues have to be dealt with in order for the control algorithm to be functional. Dynamic roll angle measurement is implemented with complementary internal sensors (accelerometer and gyroscope). Directional information is obtained through a yaw rate gyroscope which operates on the principle of resonance. To monitor the speed of the platform, a rotational sensor was formed by using a hard drive stepper motor attached to the axis of the vehicle's driving motor. The optoelectronic circuit plays the vital role to ensure the system functionality by isolating the electromagnetic noise from the motors. Finally, in order to collect runtime data, the wireless communication is implemented through Bluetooth/RS232 serial interface. The data is then plotted and analyzed with Matlab. Controller gains are tuned through numerous road tests. Field test results show that the research has successfully achieved the goal of testing the low level control of autonomous bicycle. The developed algorithms are able to balance the platform on semi-smooth surfaces.