The many advantages of harmonic drives motors such as compactness, high gear ratio, low backlash, light weight, and high torque capacity has resulted in their wide spread usage in precision control applications. However, the nonlinearities of harmonic drives including hysteresis, kinematic error, position dependent friction, and flexibility make it difficult to develop control systems that achieve precise tracking performance. In this thesis, a new approach for adaptive control of harmonic drive motors is developed using a structurally dynamic wavelet neural network to achieve accurate tracking in the presence of parameter varying friction. Furthermore, a new fuzzy logic approach is proposed for dynamic addition and removal of wavelet nodes that achieves accurate tracking using a minimum number of nodes. Experimental verification of the proposed method indicates that it can achieve precise tracking performance with a significantly smaller number of nodes than existing approaches.