If you drag a weighted rake over gravel, it forms the gravel, sort of plastically, leaving little valleys behind the rake tines. Similar phenomena occur with many soft materials and metamaterials: sand, styrofoam, oil films, wet clay, wet concrete, foamed metals, foamed waterglass, hot glass, mashed potatoes, and so on. With sufficient pressure, you can get it to happen in plastic materials like wax, aluminum, steel, and so on; lubrication and a super-hard, well-polished comb may be necessary.
Suppose that you put rounded wheels, shaped like rollerblade wheels but made of harder materials, in the tips of the comb tines. This should allow you to do this kind of plastic or pseudoplastic forming to fairly hard materials by virtue of increasing the force you can feasibly apply before the friction becomes prohibitive. Sheet-metal "spinning" is done this way, though usually with only a single handheldd wheel.
By itself, this is not a very interesting capability, because it just makes parallel lines on the surface, in fact with a fixed spacing. You can peen metal this way, but the standard ways to do that (with a hammer or with shot peening) are almost certainly better. But now suppose the comb tines are movable under precise servo control?
By varying the heights of the comb tines as you move over the surface, you can form it into any heightfield shape within some limits: you can't get inside radii shorter than the wheel radii or, in the other direction, features finer than the tine spacing; and you can only deform the surface within the plastic limits of the material. By making multiple passes over the surface you can achieve deeper deformation, especially in materials without too much work hardening. In relatively hard materials, it may be more useful to servo-control the pressure (i.e., the mechanical impedance) rather than the position of the roller; this should allow dimensional precision in the finished product that is superior to the measurement precision of your servomechanism.
Unsupported metal and paper sheets are among the easiest materials to plastically deform, but in addition to work hardening, they suffer from a limitation: it's hard to deform them into multimodal curves, because if you push down on both sides of an area, the middle goes down with them. There are several possible solutions, including supporting the sheet with a second comb of rollers, with wood, with some other deformable foam, or with an incompressible fluid, as used in hydroforming.
As with deep drawing, work hardening can be handled by normalizing or annealing the metal between stages of the process. In the particular case of sheet aluminum, which can anneal almost instantly, this can be achieved in a continuous-flow process with an air impingement oven like those used for cooking pizzas, with the air at a carefully controlled temperature a safe distance below the aluminum's melting point.
Extremely plastic materials like wet clay or wet cement might be better formed with an elastic spatula running across the ends of the comb tines rather than wheels or rollers. A similar approach might work for cutting metals or wood using an elastically-deformed cutting tool.
In all of these cases you will need some degree of FEM material simulation in the toolpath planning and perhaps even the real-time feedback system. I suspect that this has been the major reason for the great popularity of metal-cutting processes such as milling and lathe turning: although the resulting product is weaker than a forged product, the necessary planning and control is simple enough to be done by the humans' brains and massive, stiff metal frames.