When developing Bots what do you used, ie whats in the code? - Page 3

Lethal

How about on the fly navigation. bot considers options before it, ie down straight corridor with corner on left, bot hugs right wall till it reaches corner. With an added specialty looks back down the corridor where it just came from to check for enemies, then procedes not relying on a predetermined route, I know a little vague but I don't know what you guys can do with programming so I leave that bit to you's.

Pierre-Marie Baty

that resembles the "robust trace" navigation method described in the wiki, though it would not be tracing up to a particular destination. I think someone did a modified version of the HPB_bot that sorta worked that way. But it's still a very robotic-like way of moving...

Lethal

Well another suggestion could be to have bots behave like woodlice.
Mainly because I remember a project in Bio I had to do with the behaviour patterns in woodlice; turn alternation.
Could that be adotped into bots.
Maybe for bots to behave like humans you would need to begin at the basics of choice.

stefanhendriks

one of the realbot versions in the ancient times had something like that. It basicly even used 'beacons' to 'help' the bot to move from beacon to beacon to eventually get to the goal...

it was something like this:

- determine angle to goal
- face it
- check if that angle is not blocked by x radius
- if so, do a scan of 180 degrees and go for the 'most far reached' traceline.

So basicly, the bot would go for the most 'farest' traceline, as that is the most 'safe' way. (walls-wise). The theory was, when you enter a corner with 1 or 2 ways T one or L one, you could basicly have the bot turn around because of the 'farest traceline' technique. This only worked half...

why? Because due you want the bot to go somewhere (goal...) you sometimes forced a bot into a dead end, and the bot would do neatly a turn (to prevent collission with walls) and then after a few seconds get back. This is the same as you get with any path finding algorithm. Mostly noticable with First Depth Search.

Lethal

Would a Eulerian mode help?

@$3.1415rin

maybe, if we knew what you mean

... euler did quite a lot of stuff

Lethal

sorry about the ambiguatity, however you spell that hehe, I mean Eulerian bridge problem type of mode, where the nodes connected could be selected and the bot takes each connection only once and therefore may not visit the same area twice, its a mode because it will be useful when the bot is trying to hide or search for the remaining player.

Pierre-Marie Baty

That would work only if all maps were guaranteed to have an even number of entrances/exits per room. Else you would easily trap your bot in one of them. Besides, when searching for the remaining player, the bot is not searching for a static object but for a moving entity.

Lethal

Just pondering thats all.
Would a overall mind be possible?
That would make the bots play as teams but I wouldn't know how to make such a 'object'. With only one mind doing the processing it would therefore make the bots follow its commands based on the other bots encounters, well it could; I am thinking a long the lines of a hive mind controller. Worth experimenting with?
ps hope I don't sound stupid...

Pierre-Marie Baty

Yes, that's a very common technique of implementing a squad AI. But it is preferrable to delegate the control not to a hive mind, but to a squad member who will be the "officer in charge", so as to ensure the inputs are the same a human would have. Else the AI can look like it's cheating.