Bay 12 Games Forum

First of all. Maybe in real life GAs are overrated. But, I believe, they are mostly being used in the wrong way (Disclaimer: I'm a GA researcher). Regardless, in games and simulations GAs have a lot of potential and, unlike Neural Networks, are very easy to implement and use. Already some games like Europa Universalis: Rome implement simple GAs so that ingame characters inherit traits from their parents.

Toady already uses parameters to define the personalities and base skills of each dwarf so implementing a crude but functional GA wouldn't really take too much effort. Selection and fitness functions aren't needed at all because the game itself and the players do all the hard GA stuff already. So really all that Toady needs to do is just implement a way so that dwarves inherit some of their parent's attributes and Voila! DF has a working Genetic Algorithm!

The idea is simply to use a combination (crossover) of the personality and physical parameters of the parent dwarves to generate a child's parameter values instead of just randomizing them each time a new baby is born. A little mutation could be added so that each new child doesn't end up being just a clone of each parent. Also, dwarves could have initial bonuses or penalties that could be inherited, like a tendency to get sick more often or heal wounds faster or learn faster. A single parameter could act both as a bonus/penalty for two different stats (like intelligence/strength) so that a fortress doesn't end up full of dwarves having the best stats in all categories after some time.

Think of this. If a player plays a survivalist fort where dwarves have a hard time surviving they will end up being hardy and brute dwarves that survive anything and don't get stressed up but are hardly sociable or creative just after a few generations. While a player who manages too well it's fortress will end up with highly creative and productive but ultimately wuss dwarves that will succumb to a rather mild siege if the player gets caught off-guard.

[ April 25, 2008: Message edited by: Zemat ]

This is fairly minor but it would make creating fortresses in different locations more interesting. I think every biome or location should bring something different and interesting to fortress building.

[ August 02, 2007: Message edited by: Zemat ]

quote:

---

Originally posted by Jifodus:
<STRONG>Actually the design is sound. Didn't you notice the door at the bottom?</STRONG>

---

Yeah you are right, haven't noticed until now. Still without the contention walls the magma would have receded a few cells in altitude to fill the empty space. Although that isn't necessarily bad.

quote:

---

Originally posted by Asehujiko:
<STRONG>

code:

---

X
X  ########
X  #ó_____+_@@
X  ##>##### \X
X  ##>#####  X
X~~##+#####~~X
X~~||||||||~~X
X~~||||||||~~X
X~~||||||||~~X
X~~||||||||~~X
XXX||||||||XXX
XXXXX+XXXXXXXX
X____>_______X
XXXXXXXXXXXXXX

---

X=rock
~=magma
#=artificial rock
_=floor
@=dwarf
o=lever
+=door
>=stair
|=rail</STRONG>

---

There would be a flaw with this design. Assuming the railway is empty space the magma would fill the space left by the elevator and into the corridor below the flow.

Accounting for this:

code:

---

X
X  ########
X  #ó_____+_@@
X  ##>##### \X
X  ##>#####  X
X~~##+#####~~X
X~#||||||||#~X
X~#||||||||#~X
X~#||||||||#~X
X~#||||||||#~X
XXX||||||||XXX
XXXXX+XXXXXXXX
X____>_______X
XXXXXXXXXXXXXX

---

Closed chamber:

code:

---

XXXXXXXXXXXX
XXX     XXXX
XXX     XXXX
XXXX   XXXXX
XXXXXXXX===X
XXXXnnnX===X
XXóX   XXXXX
XXXXXXXXXXXX

---

Open chamber:

code:

---

XXXXXXXXXXXX
XXX     XXXX
XXX     XXXX
XXXX   XXXXX
XXXX===XXXXX
XXXX===nnnXX
XXòX   XXXXX
XXXXXXXXXXXX

---