I've been forced to take a bit of a break by my wife atm, on account of needing to do some decorating! (meaning my spare time is generally not laptop based, so i have been watching a few old Grand Prix's as research!) However, i am considering a lot of the time what's next for the game....
I may need to take an enforced break myself soon. Had a CT scan on my spine and the results are far from good. Already had two operations a long while back which gave me an extra twenty years of relief, but looks like the chooks (Aussie for chickens) may have come home to roost. Oh well, "Such is life" (attrib. Ned Kelly just before they hanged him)...
The concept of breaking the AI work up into manageable chunks also takes a lot of the load off. In some cases you may be able to break it into sequential chunks so that when an object gets its 'share' of the Vbls it only executes one chunk at a time for the next two or three Vbls in a row (to preserve continuity) then 'rests' for the next ten or so Vbls. Eg. a chunk each for 'where do I want to go?', 'what's in the way?', 'anything special happening?'. It all depends on how intense the AI will be and will probably involve a lot of 'suck it and see' to get the timing right. Obviously the physics takes precedence over everything else - collisions, etc. But there's really no limit as to how you mix and match the AI chunks to be executed each Vbl. It just needs a bit of planning and some algorithms to handle the counters and to prioritise the list(s). Strongly recommend you get the hang of linked-list structures for the data - AMOS makes it a doddle with 65,000 odd banks available and all the memory manipulation instructions. There's even some simple linked list code available in amos.library, but it's probably easier to handle in AMOS itself and not much slower. If I ever get free of the AMOS Pro bug-fixing, help updating and upgrading I'd love to get involved