This lives in large stair wells and commodious halls and just tries to stay out of trouble.
It wouldn't be a difficult extension (for an expert) to a 3D solid perspective modeller to make it a a 4D modeller, i.e., to show 2D perspective projections of solid 4D objects, and allow the viewer to "walk" around them. There is an anecdotal claim that this has been done, and that some of the people who played with the system developed the ability to see 4D objects. It would be interesting to find out if this were true.
Model helicopters can very usefully take a video camera into awkward places, but it's hard to control helicopters, so there would be a market for an intelligent model helicopter which was very easy to control and very hard to crash.
Assemblinmg Lego-Teknik has a lot of the characteristics which make assembly difficult, such as push-click fits, awkward shapes, the need for sub-assemblies, and the need to check that part-fitting operations have finished properly with the right properties (e.g. firmly stuck, or free to rotate), yet thinking about these shapes and fitting operations is greatly simplified by its modularity. It is therefore possible that one person (or a small team) could in a few years complete a system which could plan and execute Lego-Teknik assemblies. This would be an experimental test-bed for architectures and ideas in behavioural assembly robotics.
Current modelling and drafting systems won't draw you anything unless its dimensions are exacvtly specified. People often think with the aid of rough sketches in which the dimensions are simply tokens. For example, if asked to draw "any triangle" many of us will draw a triangle in which the sides are different lengths and no angle is close to a right angle. We also contruct mental maps with rough incomplete topologies and dimensions, good enough for getting around, but inexact enough that we can sometimes be surprised that certain short cuts are possible. It would be interesting to build a modeller/mapper which was capable of handling these various kinds of inexactitudes, constraints, and typicalities.