Hello groennegaard,
Quite right, and the beta-testers are the best placed people to evaluate the validity of the models. You perhaps know if a stable set of relations exists or has been established between ships parameters and movements, that would be a great point to achieve dynamic models.
ARPA and ECDIS simulation are not "far away", and better nav lights is not a major problem, surely.
AI ships are out of the question, I think. Well player ships in LAN, graphically downsized to their exterior, so as to reduce the program load, together enabling interplay of the instructors. Would it be possible to create effective and realistic AI ships, I wonder what would be the use of deck officers. Although they do not much on the bridge, they cannot be anymore missed than pilots in a plane, I am sure you agree with that (why?).
AI ships are good to populate graphically the landscape, at a distance, with elementary rules (avoiding collisions or strange behaviours), but not for close quarter situations, that are the rule in a harbor!
Although I admire the controls incorporated in the simulated bridge, I think realistic instruments need to be incorporated in different specialized panels, windows able to be transfered -ad gustum- on another display.
Nice dreams, indeed
Luc