Harmonic drives are special flexible gear systems widely used in space robots, the semiconductor industry, precision measuring devices, and military applications because of their advantages including near-zero backlash, high gear reduction, compact design, and light weight. On the other hand, they possess nonlinear transmission attributes including kinematic error, friction, flexibility, and hysteresis that are responsible for transmission performance degradation. Therefore, in-depth understanding, accurate modeling and control of transmission attributes are critical to the use of harmonic drives. There has been little research to date concerning these transmission attributes. This thesis characterizes harmonic drive transmission attributes and develops nonlinear control algorithms to enhance harmonic drive performance. The complete characterization of kinematic error in this research provides a new perspective to the understanding of the error. This thesis proposes an accurate hysteresis model in a differential equation form along with a constructive parameter identification scheme and extensive experimental validation. Furthermore, we extend a recently developed, accurate, dynamic friction model in a differential equation form to represent harmonic drive position dependent friction. Finally, several model-based nonlinear control algorithms are developed to improve harmonic drive performance. The most complex development compensates for kinematic error in presence of all other transmission attributes (flexibility, hysteresis and friction). Asymptotic stability with these algorithms is established using Lyapunov stability theory. The superior performance exhibited by these algorithms as compared to the traditional schemes is demonstrated using extensive simulation and experimental results. Thus, this thesis provides a solid foundation for performance improvement with harmonic drives as well as with other systems sharing one or more of the transmission attributes.