Re: FOV/scent-based pathfinding

On Apr 5, 5:55 pm, Brigand <markash...@xxxxxxxxxxx> wrote:
Yeah, I thing Mingos hit it on the head. Designing an intricate,
detailed, completely bug-free algorithm for tracking may be quite an
acomplishment, a step towards realism which should be applauded...
but, unfortunately, without expressly telling the player what's going
on, they are going to be largely oblivious to it. The average person
isn't going to consider why a monster moves the way he does, only in
how he moves.

I also use a form of Ray's algorithm, and it was simpler to implement
than A*. It's not supposed to simulate realistic bloodhound-like
monsters -- at least I don't use it to simulate that -- but rather
monsters who see you and chase you intelligently, whom you can't lose
by ducking around a corner. In fact the whole "scent" metaphor is a
bit misleading; since when after all does scent depend on line of
sight? I'm afraid people see the word "scent" and assume it's a
misguided effort to import a pointless degree of realism. It's not;
it's a simple, low-overhead, purely functional algorithm to make
monsters chase you.

95% of the time you're probably better served
(especially in computer calculating resources) by If Player_X >
Monster_X then Monster_X=Monster_X - 1 type tracking, with more
complicated pathfinding only coming into play if the creature gets
stuck.

Except that you do need the more complicated pathfinding algorithm to
take over in the 5% of the time when the player ducks around a corner,
because if you don't have it, the player is going to start ducking
around corners a lot more than 5% of the time. And if you do have it,
then you have to code the alternative pathfinding algorithm, code a
method of detecting when a monster needs to switch over to the more
complicated algorithm, embed that information into the monster so it
remembers from turn to turn that it's following the complicated
algorithm, etc. Ray's algorithm, on the other hand, is a single
approach for having monsters chase you (around corners *and* in line
of sight) that avoids the need to run dual pathfinding algorithms in
parallel. It's a lot simpler. Of course if you want your monsters to
wander, flee, etc. in addition to chasing the player, then you need
parallel algorithms, but one fewer is always easier.
.

Relevant Pages

  • Re: traffic flow in roguelikes
    ... algorithm for handling this traffic flow. ... a gradient for the monsters to follow. ... all the tiles connected to the goal would have a ... Monsters always move from a higher value tile to a lower value tile, ...
    (rec.games.roguelike.development)
  • Re: Monster Patrol
    ... A minor caveat would be that some corridors are left ... Routing nodes are more data structure than algorithm, ... You can keep track of where the player ... monsters intelligently protect their property. ...
    (rec.games.roguelike.development)
  • Re: Compressing your levels
    ... resizing of bitmap graphics: ... the algorithm (with a modified function ... Angband, as you know, has one of the largest dungeon sizes. ... are packed with monsters, etc. ...
    (rec.games.roguelike.development)
  • Re: Crossroads - when do you recode, and when do you start over??
    ... something I could never achieve with an algorithm. ... Hmmm... ... Monsters act how I want them ... universally - and you can easily make it stupid by setting a limit on ...
    (rec.games.roguelike.development)
  • Re: Two EPR questions
    ... >>some Psiwhich is not spatial but defined on the configuration ... does not describe some state of reality but our knowledge about ... >>algorithm has been guessed correctly. ... > sense" which is inevitably reached by proponents of realism ...
    (sci.physics.research)
Loading