AI Playerbot

[Christian Schamara]

sorry not support here for mangosr2 check the mangosr2 forum for support thanks

[iceymann]

just wondering has anyone tried this on mangosone or mangostwo?

[Nykaen]

Tested bots with 3.3.5 server - worked like a dream, I was even successful to do several dungeons.

I was wondering does playerbots work with 4.3.4 core?

[kutyirigo]

is it possible to somehow compile this playerbot with the recent MangosZero core? or we have to wait till they update the current playerbot to be compatible

[Unkle Nuke]

AI Playerbot was originally developed for MaNGOS Two (3.3.5a WoTLK), but ike decided to work on it for other projects like R2 and Cmangos.

There are branches in his repo for MaNGOS Zero, One, and Two. However, ike never went past 3.3.5a.

We would love to see some of you in our community take up this project and develop versions for MaNGOS Three and Four, plus keep the code for Zero, One, and Two updated.

So who is up for the challenge?

[PsiDraconis]

I have successfully merged ike3's original Playerbot-AI code for MaNGOSZero with the most recent release of the core code (Rel 19, as of now), and the result is a functionally playable product. There are still fairly major glitches, some because of changes to the way the source code was written (e.g. bots can loot corpses, but not gameobjects), and some simply because it needs polishing (e.g. bots often have trouble turning to face an opponent that is attacking them, and therefore cannot attack in return). I will be working on ironing out these bugs as my (limited) time permits, but I am more than happy to open the floor to anyone who wants to help. I will get around to setting up an "Issues" page on Github, but for the moment, please don't contact me simply to report a bug. Chances are I know about it already, and will try to get around to it when I have time. At the same time, I am happy to accept help and suggestions.

Get the code at https://github.com/PsiDraconis/MaNGOSZero-AI, and find instructions to compile elsewhere on this forum (it's exactly the same method as MaNGOSZero Rel19).

For the basics on actually using Playerbot once compiled, see ike3's very helpful page at MaNGOS-Bot

Here's to bug-squishing!

PsiDraconis

[krill]

I have successfully merged ike3's original Playerbot-AI code for MaNGOSZero with the most recent release of the core code (Rel 19, as of now), and the result is a functionally playable product. There are still fairly major glitches, some because of changes to the way the source code was written (e.g. bots can loot corpses, but not gameobjects), and some simply because it needs polishing (e.g. bots often have trouble turning to face an opponent that is attacking them, and therefore cannot attack in return). I will be working on ironing out these bugs as my (limited) time permits, but I am more than happy to open the floor to anyone who wants to help. I will get around to setting up an "Issues" page on Github, but for the moment, please don't contact me simply to report a bug. Chances are I know about it already, and will try to get around to it when I have time. At the same time, I am happy to accept help and suggestions.

Get the code at https://github.com/PsiDraconis/MaNGOSZero-AI, and find instructions to compile elsewhere on this forum (it's exactly the same method as MaNGOSZero Rel19).

For the basics on actually using Playerbot once compiled, see ike3's very helpful page at MaNGOS-Bot

Here's to bug-squishing!

PsiDraconis

Nice, ty

[madmax]

Thanks for this, I've passed this along to antz

[k109]

PsiDraconis - you rock! This is exactly what I was looking for. I built the release a few hours ago in VS 2010 and it's working great (although I did have to copy in the Playerbot folder, in the release configuration it was not there). The bots seem to work very well. A few minor bugs, but I'll dig through the code and see if I can help fix some of them.

Thanks again!

[cabfever]

Could mangos implement this as a submodule?

[antz]

It is definitely something worth looking into. would need to operate like ahbot or like SD2/Eluna modules.

But can be discussed !

[Xenithar]

I would like this as a sub-module also. It would be nice to have a few random AI players in the world to group with or fight against from time to time.

[antz]

Once we have the Rel20 branches out.... This would be awesome to be picked up by someone and implemented

[Foereaper]

Should be easy enough to do

[Foereaper]

After discussing this, we're pretty interested in implementing this as a submodule, however it would be nice to have the original developer (in this case Ike) on board with it as well! Not sure whether Ike checks this forum or not? Doesn't seem like he has been actively developing the bots on his github for a while..

[Unkle Nuke]

Last I checked, ike3 abandoned Cmangos and now supports only Trinity and Mangos R2 with his new mangosbot project which seems to be a continuation of Playerbot-AI, but he hasn't made any commits to it in roughly 3 months.

Regardless, the entire reason I started this sub-forum was to drum up interest in creating a MaNGOS-compatible fork.

Still waiting.

[antz]

I have PlayerBot-AI on my hit list of new features for Rel21

[Xenithar]

The current build in R21 is decent for basic questing and I am becoming interested in the project and furthering the AI. I have been looking over the code for a few days and am stumped as to where the commands are. The code could use some serious commenting. I believe if I can pick it up I can expand what this does and am actually excited to do just that. I'll keep looking at it and testing.

[madmax]

Playerbot system and Ahbot are all planned to be moved to there own daemons and will likely receive massive updates/optimizations at the same time.

[MENTION=2]antz[/MENTION] could probably give you more detail on whats planned with those.

[Foereaper]

The current main issue with the playerbots is the spell and gear availability as well as zones etc. The playerbots in zero are based off of 3.3.5a and the play style in said expansion. This causes some of the logic to fail, increasing the resource use and stability for bots as a whole (An example is Mage bots, they constantly attempt to cast conjure food, lagging the server out). This can easily be fixed by someone revising the logic and AI for each class, however it's not something I ever got around to do.

On that note, I am completely against separating the player bots specifically from the core. The way the daemons communicate is through tcp/ip, which would cause a massive increase in network bandwidth usage, especially between the world and playerbot daemon. It's not feasible to stress the world daemon more than necessary, especially when the bots are literally utilizing the packet handler and player functions directly.

AH bot on the other hand doesn't need access to the core what so ever and could easily be separated. The only thing the AH bot would need to do is monitor the AH tables in the database, nothing else is really needed.

[Xenithar]

I agree with Reaper, moving the playerbots to a daemon would really increase network latency and, on virtualized (read; modern) setups that would increase CPU usage on the virtualized hardware. Keep the bots in the core. I will start looking at the AI. My big beef is the commands not confirming success. For example "~co +fire,-frost" results in my mage doing the correct thing, but he does not tell me his style changed. I also cannot list what the existing orders are. I plan on fixing that, but the chat parser and commands are strung out all over the place and they're not commented.

As for the AI however, I know right where those files are and can look into understanding them and updating them. I will begin that this weekend. I really like the bots and have been going in just to mess around with them a lot lately, and would love to improve them. If you have any advice, feel free to post it. Otherwise I'll be doing this starting Saturday night.

[Foereaper]

What I would personally do before starting to work on the AI, is note down ALL the spells for the specific class you are going to work on, strip everything out and start fresh.Not only would this give you a proper understanding of how the AI is set up currently, but you'd also instantly root out all the issues with the current AI, be it wrong spells or bad logic compared to 3.3.5a.

If you look at this original commit, it shows you all the files edited for the player bots. Implement Playerbots · mangoszero/server@ee33683 · GitHub

Notice the changes to Chat and ChatHandler specifically, where all the commands are parsed (Exactly like all the other core commands). This is where you'd add your confirmation/debug messages.

[Xenithar]

Alright, I may do that. It would indeed make it easier to start fresh on the AI, and if I focus on one class at a time, I could probably get something passable in a short amount of time. I guess the only thing left to ask is what everybody agrees on for how each AI should react. For example, I remember playing my mage solo plenty back in the day. I'd run into a group of mobs, root with frost nova, blink, and cast something (cannot remember the name) which did AoE damage. This would not be good in a group, obviously. In group play I normally stayed back and hurled fire and ice at things.

If there are suggestions for how a class AI should work grouped, post them here. I am digging into the code now, but will be leaving in an hour or two for about four hours.

[Lidocain]

Deleted

Edited March 22, 2017 by Lidocain
Removing a link to ac-web including a Paypal link.

[The0dark0one]

On 1/30/2016 at 6:53 AM, Xenithar said:

Alright, I may do that. It would indeed make it easier to start fresh on the AI, and if I focus on one class at a time, I could probably get something passable in a short amount of time. I guess the only thing left to ask is what everybody agrees on for how each AI should react. For example, I remember playing my mage solo plenty back in the day. I'd run into a group of mobs, root with frost nova, blink, and cast something (cannot remember the name) which did AoE damage. This would not be good in a group, obviously. In group play I normally stayed back and hurled fire and ice at things.

If there are suggestions for how a class AI should work grouped, post them here. I am digging into the code now, but will be leaving in an hour or two for about four hours.

I was thinking about playing around with this as well, did you manage to make any progress?