Development of Clusterwheel Balancing Robots and Methods for Improving Performance of Wheeled Inverted Pendulum Machines

There are a number of robotic locomotion architectures that have been explored over the last decade and more. As software and electronic control capabilities have steadily advanced, the design of complex mechanical architectures and sophisticated control became feasible. Presently there exist many robotic architectures that use legs, wheels, treads or a combination of these for locomotion. There are also many architectures that have been designed specifically for safe operation in human environments. However at present there exists no hybrid architecture that combines the energy efficiency and reliability of wheels, human friendly compliance of dynamic balancing and terrain capabilities of legs. In this dissertation we describe the design and exploration of a new class of robots called "clusterwheel" robots that seek to combine these advantages in a single platform.;This dissertation represents a first step in studying the design, fabrication and feasibility of "clusterwheel" architectures. Our objectives are three fold: First we describe how to achieve better dynamic balancing on two wheels by studying aspects of robot construction and integration that affect balancing performance. We describe tradeoffs between stability, energy efficiency and vehicle speed for subcomponent characteristics such as tire damping, motor gearing and current/voltage control of actuators. These non-linearities due to sub-component properties pose a significant challenge to system performance and especially so for zero-moment balancing machines such as wheeled inverted pendulum machines.;Next we explore the geometric architecture of clusterwheel robots and use results from the analysis of system dynamics to inform our selection of cluster geometry. In investigating clusterwheel architectures we demonstrate analytically and by simulation the relative stability of the three wheel cluster over the two wheel cluster. We then describe the design and construction of "Charlie" a clusterwheel balancing robot and demonstrate tele-operation of the robot while climbing simple obstacles.;Finally we explore electromechanical design of clusterwheel robots for autonomous two to four wheel transitions. Our investigations into cluster stabilization strategies reveal the reduced control authority of the cluster actuator. These results guide our design for a new clusterwheel robot "Charlie V2" emphasizing the power of the wheel motor over the cluster motor. Using this robot we then demonstrate open-loop autonomous two to four wheel transitions.;Our results represent advances in understanding the dynamics and control of wheeled inverted pendulum robots and clusterwheel robots. By showing the importance of selection of tires, motor gearing and motor control schemes we demonstrate the importance of system integration in wheeled inverted pendulum performance. And by showing the importance of cluster geometry and high-power actuator requirements in clusterwheel robots, we have laid the foundation for future advances in the design and control of these robots.