Has anyone every done any work around setting up trade routes as optimisations of linear programs?
No, but I did spend three years developing transport models to study the economics of congestible transport networks, particularly the streets of Australian cities.
As far as I can see you don't need to use the Simplex Algorithm when you are dealing with space transport because the links are not capacity constrained and the node capacities are expandible by investment. You can just assume that the starports and ships will expand to market equilibrium. It's just a matter of generating volumes on OD pairs and then assigning them to routes. Or were you thinking of linear constraints to handle multilateral balance of trade?
You should, however, look into the Four Step Model for transport modelling (trip generation, mode split, destination choice, route choice). ou won't need to iterate it because your links aren't congestible, which means utilisation won't affect the skim matrix: it's just a matter of finding shortest path lengths, and the Simplex Algorithm is overkill for that . (Mind you, the algorithmic efficiency is not so bad with today's desktop resources, and it is at least simple to use Simplex for everything. But back in my day I was using a 90 MHz 486 DOS box, and iterating my models up to thirty times to get convergence.)
You are soon going to discover that trade economists use gravity models, sometimes fitting the exponent econometrically. But that is just a hack: there is no economics behind gravity models. I would suggest that you try out multinomial probit modelling, which has foundations in discrete choice theory. If you treat each planet as one of its own trading partners (which is actually simpler) you ought to be able to generalise out the trade/GWP value as an endogenous prediction of the model.
Or were you thinking of triangle trade and the back-haulage problem?
). All that kind of stuff. In effect, where are the dark alleys of the subsectors. 
