Bots and Simulations > Bot Tavern

The Wraithbot

<< < (2/3) > >>

Martian:
I think that we should make the queen build a nest with tiny bots, like proposed in this thread. They would be fixed.

They would shoot any enemy bot with 1/3 of their total energy in .shootval.

Then we wouldn't have to restrict the preybots to a part of the screen, any preybots coming close to the nest will be killed.

I think we should also make the preybots run away terrified whenever they see these guardians. Perhaps make them store the x,y pos of the nest in their out/in and racial memory, and avoid the area around it.

We could perhaps later on make each queen store a random number in the racial memory. This would make two different queen's bots attack each other. Then we could add another bot the Bodyguard. They're armed very well (venom, poison, shell etc.). Their work is to defend the hive from enemy wraithbots.

Perhaps we could make another bot which job is to store energy. Then whenever a bot runs low on energy, especially the minebots, they could come and get some from the store with ties or shots.

Just throwing around lots of ideas here. Tell me what you think.

Martian

PS. I'd like to help, Could someone please deal me in?

Numsgil:
There are a couple threads scattered around.  Mostly no one took initiative and wrote it.

fulizer:
or make a pondmode only bot that uses the bouancy command to make them float away never to be seen again to be killed by something eles

rsucoop:

--- Quote from: Numsgil ---There are a couple threads scattered around.  Mostly no one took initiative and wrote it.
--- End quote ---

One problem I see is that in order to write such a bot, a lot of important variables must be considered. 1- Importance of Genetic Presurvation, 2 - Importance of how many bots it would protect and harvest, 3 - with such a level of randomness, a delgene code in the inactive genes would destroy the dna at the end of the life cycles. So a very large number of variables must be kept and stored into x (number of protected bots  + any enemy info + conspec info + all ties and their refs). The number of required variables to be efficient, say 2 veggies tied + 1 conspec + 2 enemies * (say, refvel, refnrg, refpain*, *dnalen, *vtimer, xpos, ypos, timer since last seen (max 10 seconds - giving a 10 second recycle if no other 'new' bot) for memory refresh purposes and possible hibernation when nothing is within 100 distance for say 3 seconds. On top of that you would have to design various life stages to the bot to be truely effective in a simulation, such a step can help prevent mutations that are really bad, from destroying the bot. To add to this, you must also allow for other varients to exist for the bot, meaning that conspec recognition is key, so at least 2 outs must be used, along with dna some other reference to compare with. All of this must be preserved and would require a very large amount of planning, one required outside the internet, with diagrams and what not.

And I think such a bot would need every variable to have some sort of base standard to revert back to when they become a value not within the bots perameters (perfect for antivenom/infoshot/antipoison/antivirus).  Then of cousre you would need a very effective antiviral gene to make sure the bot doesn't lose all this work by some effective gene deleter.

I think I would take the challenge if someone would be willing to collect a link to every descent worthwhile discussion on the topic.

Numsgil:
Most bot designs seem to either require 10 custom variables or 10000.  It's sort of a weird issue I'm not sure how to address.  1000 memory locations is either way too many or way not enough for most uses.  Not to mention the practical issues of using huge numbers of custom memory locations and keeping it all straight.

Navigation

[0] Message Index

[#] Next page

[*] Previous page