RE: Thought process of a colonizer

["Stanley Sutton" <sutton at t-surf dot com>]

Comments:
With such a steep gradiant on the movement reducing the potential value,
you don't have to search all squares, just the ones withing n moves of
the starting point, where you can calculate n based on whatever will
reduce the value to colonize to less than some value.
 
Don't do the evaluation every turn, just do it  when you start, or
arrive at a destingation.  I'd add a large bonus for being on a
colonizable square to keep from having "wandering" colonizers.
 
Add a terrain property which precaluclates most of this, as it should
only change if terraforming or exploration goes on.  You can cheat, and
know the value at board generation time, or be more realistic, and
calculate it as new territory becomes available.  

	-----Original Message----- 
	From: Lincoln Peters [mailto:peters2000@mindspring.com] 
	Sent: Fri 21-Jun-02 20:28 
	To: Xconq mailing list 
	Cc: 
	Subject: Thought process of a colonizer
	
	

	After reading James's e-mail, I've been trying to think of a way
that
	the same "Colonizing" plan could work for both terrestrial and
space
	games, and I came up with the following thought process that the
AI
	should probably use.
	
	(For clarity, I'll assume that the unit that the colonizer will
build is
	a city.  Footnotes for each section appear underneath that
section.  I
	also assume that if any value referred to is undefined, it will
appear
	to the algorithm as a zero.)
	
	--- BEGIN MECHANICAL THOUGHT PROCESS ---
	
	1. Consider the terrain around the colonizer.  It should
consider every
	tile that is known to the side, regardless of where it is (more
on where
	it is later).
	1a. If the known terrain does not include a radius that equals
of
	exceeds that of the unit that the colonizer will build (e.g.
terrain
	around colonizer is known 2x2, but reach of a city is 3x3), call
	subroutine A.  Otherwise, go to step 1b.
	1b. Evaluate every known cell for its value to a city.  Call
function B
	to determine the value.
	2. Given the reach of the city, determine the ideal spot(s) for
a city.
	To do so, process each known tile, and add the scores of those
tiles
	that would be in a city's reach if built there (use different
data
	tables so that nothing weird happens).  The ideal spot to build
should
	be the tile that gets the highest score by this test.
	3. Scan through the list of possible spots to build, and delete
any
	options that match any condition in function C.
	3a. If no spots remain in the list, call subroutine A (i.e.
explore
	farther), then go back to step 2.
	4. Multiply the score of each tile by s/(d+1), where s is the
speed of
	the colonizer (i.e. how many tiles it can traverse in 1 turn),
and d is
	the distance from the colonizer to that spot.*
	5. The tile with the highest score is the one that should be
settled.
	If there are no problems with that spot (e.g. it's already
occupied by
	an enemy city), go for it!  Otherwise, go to step 5a.
	5a. If, for some reason, it can't be settled, pick the next-best
spot
	and try for that one.  If necessary, keep going down the list
until an
	open one is found.
	
	* This penalty for distance may be too steep.  The formula might
need a
	few constants attached in order to make more distant spots less
	undesirable.
	
	
	A.
	1. If it is possible to order a nearby non-colonizer to explore
the edge
	of the known map and thereby gather the necessary information to
	determine the area's value as a spot to build, do so.  Otherwise
the
	colonizer should do the exploration itself.*
	
	*    Perhaps Brownian motion would be more appropriate as part
of the
	"Exploratory" plan than as part of the "Colonizing" plan?
	
	
	B.
	The properties of the tile should be considered as follows:
	1. Look at the unit-consumption-per-size of the city.  Assign
each
	material listed 1 point for each unit needed per size.
	2. Look at the unit-consumption-to-grow of the city.
	Assign each material listed 0.5 points for each unit needed to
grow.*
	3. Look at the unit-consumption-per-cp of every unit that the
city will
	be able to build.  Assign each material listed 0.1 points for
each unit
	needed for each CP for each unit, then divide by the unit's CP.
Add the
	final result to the tile's score.
	(might need a few more tests in here)
	4. Re-evaluate from step 1 using what-if conditions based on the
city's
	possible facilities.  An example of such a condition to evaluate
might
	be "What if I had a granary?".  These conditions should not be
mutually
	exclusive (e.g. allow "What if I had a granary AND a
university?").
	Multiply the result point value of each test by 1/x (where x is
the
	number of what-if's processed), and add the final result to the
total
	score.**
	5. Return the final score as the value of the tile.
	
	*   Maybe instead of using 0.5 or some other constant, the value
should
	use some kind of decay curve so that initial growth is more
important
	than later growth.
	** This may be the key to writing a better "Improving" plan!
	
	
	C.
	1. Is the area clear of cities whose reach overlap that of a
possible
	city in this spot (i.e. would any other cities would try to use
the same
	terrain as ours)?*
	1. Look at the materials that are listed in the city's
base-consumption
	and consumption-per-size.  Also, if necessary, consider the
materials
	that are in the city's base-production (adjusted for
productivity, if
	necessary).  Can the spot in question produce those materials in
at
	least the amounts consumed minus the city's own production?
	2. If a city is built here, will the terrain provide all of the
	materials needed to build any new units that the city might
build?
	(Don't worry about how much is produced; just that it's produced
at all)
	3. Regarding question 2, what size must the city be in order to
produce
	all of those materials?  Will the terrain support a city of that
size?**
	4. If any of the above questions were answered with a "no",
consider the
	spot unsuitable for building and delete it from the list.
	
	*  Perhaps if there is a hostile city overlapping the area, it
should be
	considered whether or not to attack.  But for now, I'll focus on
	colonization, not combat.
	** It might be worth considering the impact of facilities such
as
	granaries, but that entails more complexities than I'm willing
to worry
	about at this time.
	
	--- END MECHANICAL THOUGHT PROCESS ---
	
	I hope that this all makes sense.  The only problem I can see
with this
	setup might appear in games like "fantasy" or "postmodern",
where cities
	are non-advanced, but there are colonizer units that can build
them.
	They would probably do best to just have their own algorithm
that would
	only consider the values in the productivity table.
	
	So, does it make sense to anyone?  Any questions or comments?