Microscale additive manufacturing (µ-AM) processes are a class of manufacturing processes used to fabricate micron-sized structures in a sequence of direct additions of materials as instructed by a digital file, as opposed to the lithographic patterning and subtractive etching used in traditional microscale manufacturing. Despite being sophisticated, numerically controlled tools, material addition is an open-loop process which requires continual user intervention to heuristically tune process parameters. In addition, the layer-to-layer dynamics in µ-AM are not well understood. This dissertation investigates layer-to-layer dynamics from a system identification perspective. This work defines a class of input signals, system identification algorithm for µ-AM modeled as a discrete repetitive system, and the experimental protocol to empirically the plant model and validate the model for a different input signal. A case study applied to the µ-AM process electrohydrodynamic jet printing demonstrates that the identified model from a training set is extensible to a validation data set, with less than 4% error between the system identification of the training and validation data sets. Models describing the dynamics of the µ-AM enable the design of model-based controls. This dissertation details the first experimental demonstration of a run-to-run feedback algorithm termed Spatial Iterative Learning Control (SILC), a framework enable robust, auto-regulation of sensitive µ-AM processes. This work demonstrates that SILC enables us to autonomously fabricate complex topography structures with as small as 5µm x-and y-axis resolution and ∼113nm feature height accuracy, without any heuristic tuning by a user. Lastly, SILC can be designed to be robust to system faults, as demonstrated by the ability to recover from both an actuator and sensor error in two iterations.