Ballbots are a family of mobile robots that ride on a ball, or spherical wheel. Many ballbot drivetrains have been developed, while the application of ballbots in our daily lives is still rare. There lack design principles to build the ballbot drivetrain for practical applications exploiting their unique features:​ omnidirectional maneuverability and dynamic stability. In addition, there lack investigations into the physical human-robot interaction of ballbots, which could provide an intuitive interface to control the maneuverability of ballbots via physical interactions. The work in this dissertation investigates the design, modeling, and control of a ballbot drivetrain developed for payload-carrying and, more importantly, human-riding applications. Two drivetrain prototypes were developed. The first prototype utilized a popular drivetrain design and a novel cascaded model-based controller, which achieved human-walking level agility in a single translational direction. However, it could easily lose dynamic stability in other directions due to omniwheel-spherical wheel slip. To address this issue, a second study was conducted to evaluate the effect of drivetrain design variables on the torque generation capability and contact stability of the ballbot, and to identify principles for ballbot design. A design optimization problem was formulated and solved to obtain optimal drivetrain configuration with three and four omniwheels. Following the optimal design, the second prototype showed improved contact stability and achieved human-walking level agility in all tested translational directions. The main piece of this research is a human riding ballbot, Personal Unique Rolling Experience (PURE), developed based on the second ballbot drivetrain and a chair module that measures the torso dynamics and physical interactions. To compensate for the difficulty of PURE riding in a seated position, controllers based on impedance control and admittance control were developed to enable effective hands-free control of PURE via torso motions. Human subject tests were conducted to determine the most effective hands-free control scheme and the feasibility of riding PURE for various indoor navigation tasks using torso motions. In addition, to accommodate parameter uncertainties and externally applied interactions to the ballbot, a novel control scheme was developed with an offline control gain personalization module and an online interaction compensation module. Using this control scheme, we demonstrated simple shared-motion control tasks of PURE including idle-keeping and speed-limiting with and without this interactive hands-free control scheme. This feature allowed the ballbot to provide physical interactions back to the rider to ensure system safety and reduce the proficiency requirement of riding PURE.