Multiscale Granular Stacking (MSGS) is a novel additive manufacturing process reliant on in-situ rocks and grains for constructing communications towers, radiation shields, parabolic dishes, habitats, and other elements of lunar infrastructure that are necessary for a high-functioning and safe lunar base camp. A fully functioning MSGS system operates on the lunar surface with collaborative rovers stacking un-processed rocks and grains to construct predetermined structures from small amounts of pre-processed binder materials, power, and communication capabilities available during the earliest phases of lunar base camp development. This work details attempts to bring multivariable control theory to bear on additive manufacturing by using feedback on overall build geometry. Modeling these fabrication process dynamics as a discrete-step linear system allows for the tuning of parameters such as build speed, surface finish, and contour smoothing while providing the opportunity to leverage control theory for determining system convergence, steady-state error, and overshoot of desired build height. Finite-element models composed of volumetric elements that represent the MSGS builds assess the material properties in a Monte Carlo approach. A structure may require stacking millions of grains, which demands extensive computation. This work explores two model reduction methods: 2D Discrete Fourier Transform state elimination and Balanced Realization model reduction. Both methods show significant reduction in dynamic system dimensions while maintaining the desired build shape.