The problem is complex in that an ore is not a continuous surface, but rather something that is discontinuous in space, and in concentration, so that knowing a rough position within the ore field, one needs to determine the best path to excavate/treat the best increment of the surrounding space so as to achieve an optimal extraction of the surrounding ore. (this is a remote sensing problem including lidar, GPR and other existing technologies inputting to an assessment algorithm.)
Having established such a direction, the next step is to disaggregate the material in that direction, so that the mineral can be separated from the relatively no-value host material (gangue). This can be demonstrably achieved (one solution being high intensity cavitation which attacks along grain boundaries followed by, with some ores, a simple size separation, with others a more complex division, but in both cases at the mining machine, so that only the valuable material is transported back to the subsequent processing plant, while the residual waste is incorporated into the underground roof support mechanism).
Surrounding gangue must then be removed to provide access for the equipment, as it advances to mine deeper along the vein of value. This can be removed with a simple crack-and-split technique (see the removal of rock from under the Gateway Arch in St Louis). This material does not need further processing but likely can be intelligently broken into suitable sized pieces to occupy mined volume and provide support to the mining forward.
The extracted valuable components are moved, through piping (to minimize impact on the feed tunnel environment), to a supply shaft and thus to the surface, where the valuable minerals are refined into a commercial product at a much reduced transportation and processing cost.