Design and test of a computer-stabilized unicycle

Abstract

This research investigates the stabilization of a one wheeled vehicle by means of active feedback control. The control methods of a human riding a unicycle are investigated first and a dynamic model which closely emulates the process is derived. A one wheeled robot with mass and inertia propeties similar to those of a young child was constructed and used as an experimental vehicle for testing various control algorithms. The research addresses aspects in the fields of robotics, artificial intelligence and modern digital control, but rather than specializing in any of these fields, it strives to combine these disciplines in a unique application where the interaction of these fields can be studied. An underlying approach of this research was to not only design but also evaluate control system performance in a laboratory environment without incurring large financial expenses. The robot has all its electrical and computational power on board, with the ability to receive commands from a radio transmitter to change its direction and forward speed. A linearized model was derived and optimal control systems to stabilize the vehicle were designed and simulated. An investigation into using accelerometers for detection of the deviation from vertical by measuring the specific force on the robot frame, was conducted. We found that this resulted in unacceptable closed loop system robustness. Theoretical and physical explanations for this phenomenon are presented as well as experimental results to confirm the extreme sensitivity of the design to these sensors. We show that accurate sensor information on the unicycle's orientation with respect to vertical facilitates the design of closed loop control systems with good stability and robustness characteristics. Such a control system for the longitudinal dynamics of the unicycle robot was demonstrated experimentally. The sensing, actuation and control abilities of a person riding a unicycle are compared with those of a computerized robot performing a similar task. We propose that this research and the test vehicle form the basis for theoretical and experimental studies into the application of nonlinear, robust and adaptive control systems techniques for unstable systems.