Links people who are developing bots might find usefull

[!@^^@!]

Link, links and more LINKS!

Well it seems to me there is a certain lack of links for people who just started developing bots like me, note though: most of these links does not include any code at all but is rather theoretical stuff instead and yes many of these links have been posted before but are scatered around the site...

Path finding:

Game/AI: Fixing Pathfinding Once and For All

Amit’s A* Pages

Tanis - Articles - Path finding in C#

recastnavigation - Project Hosting on Google Code

MPQNav Public Blog


Behavior:

http://www.behaviorengineering.org/d...tation-1.0.pdf

http://www.behaviorengineering.org/d...aresubmit2.pdf

Behavior Trees for Next-Gen Game AI— AiGameDev.com
Version that does not req. a free signup is in 3 parts (first version includes slides to make it easier to understand):
Behavior Trees for Next-Gen Game AI (Video, Part 1)— AiGameDev.com
Behavior Trees for Next-Gen Game AI (Video, Part 2)— AiGameDev.com
Game AI for Developers — AiGameDev.com

My Liner Notes for Spore/Spore Behavior Tree Docs - Chris Hecker's Website

Brainiac Designer

http://www.mmowned.com/forums/c/2327...your-bots.html

On Finite State Machines and Reusability— AiGameDev.com

The Gist of Hierarchical FSM— AiGameDev.com


Injection/Hooking:

EasyHook - The reinvention of Windows API Hooking

http://www.mmowned.com/forums/wow-me...r-library.html

Detours - Microsoft Research


Programming languages:

Getting Started with C# — Developer.com

C# Station: C# Tutorial Lesson 01 - Getting Started

How Do I in C#

Cprogramming.com: How to Get Started with C++

C++ Language Tutorial


Marshaling:

http://www.codeproject.com/KB/cs/marshaling-1.aspx

http://www.codeproject.com/KB/cs/marshaling-2.aspx

http://www.codeproject.com/KB/cs/marshaling-3.aspx


Misc:

How to find stuff

http://www.mmowned.com/forums/wow-me...ple-stuff.html


Okay this list is far from done, therefor if people post links in this thread i'll add them... Maybe we'll end up with a post that could help people(including me) before they ask a stupid question and get a kick in the bawls of Apoc/Cypher

[amadmonk]

I've been wanting for a while now to do a piece on the difference between planners and scripted behavior (hardcode/FSM/BT's) for bot AI. I'd highly recommend that folks dig into this a bit if they want their bot to be more "clever." A true planner is a beautiful thing, but quite heavy just for a game bot (Apoc didn't you say you wrote a GOAP a while back?). A BT is probably the least "scripted" of the scripted behaviors family -- you still need to plan out in advance all of the logic paths, but the individual components are composable and easily changed.

The one tech that I really wouldn't recommend anyone spend a lot of time on for game AI is FSM's -- they've fallen out of fashion in the game design world, for a reason.

[!@^^@!]

Yep FSM's seems like the way you only might go the first time because it's simpler to bruteforce to work...
But depending on how dynamic you want your bot to be i think a behavior tree is a little overkill for example a bot that use breadcrumb pathing and just grids would be ideal for a FSM because it's not complex enough to justify a behavior tree... While a quest bot could push a behavior tree to it's limits and give a awesome result...

I really find these forms of AI very interresting because the amount of dynamic within your rules is truely awesome

[Screamer2010]

Cool, Thanks for the excellent info

[nitrogrlie]

I've been wanting for a while now to do a piece on the difference between planners and scripted behavior (hardcode/FSM/BT's) for bot AI. I'd highly recommend that folks dig into this a bit if they want their bot to be more "clever." A true planner is a beautiful thing, but quite heavy just for a game bot (Apoc didn't you say you wrote a GOAP a while back?). A BT is probably the least "scripted" of the scripted behaviors family -- you still need to plan out in advance all of the logic paths, but the individual components are composable and easily changed.

The one tech that I really wouldn't recommend anyone spend a lot of time on for game AI is FSM's -- they've fallen out of fashion in the game design world, for a reason.

Didn't we just talk about this yesterday? Here is a link to some good GOAP information: Goal-Oriented Action Planning (GOAP). The source code is a bit over-the-top as it is for the actual F.E.A.R. implementation which is specific to their data information for characters, but some of the papers should be sufficient to start people down the right path of proper goal and action class designs. A* is A*... what's new?

[Apoc]

Didn't we just talk about this yesterday? Here is a link to some good GOAP information: Goal-Oriented Action Planning (GOAP). The source code is a bit over-the-top as it is for the actual F.E.A.R. implementation which is specific to their data information for characters, but some of the papers should be sufficient to start people down the right path of proper goal and action class designs. A* is A*... what's new?

Due to the fact that we're forced into a reactive AI system, GOAP is way over the top as far as deciding which AI path to take.

You would literally need to implement every possible change which would be a giant pain in the ass. Not to mention; a world 'state' system to determine what's actually going on, and what we 'want' to happen is near impossible with certain aspects of WoW. (PvP mainly)

You can only plan so far ahead when writing logic for your bot. The rest is mostly reacting to events.

[nitrogrlie]

Due to the fact that we're forced into a reactive AI system, GOAP is way over the top as far as deciding which AI path to take.

You would literally need to implement every possible change which would be a giant pain in the ass. Not to mention; a world 'state' system to determine what's actually going on, and what we 'want' to happen is near impossible with certain aspects of WoW. (PvP mainly)

You can only plan so far ahead when writing logic for your bot. The rest is mostly reacting to events.

Right, but there is more than one way to "react" to an event... hence a goal and multiple actions that could satisfy it under different trigger conditions. Currently I'm ignore "pathing" all together for movement. I just want to put together a 2v2 arena decision system. I started with a FSM but problem is that it gets complicated fairly fast... especially when I start dealing with fuzzy predictive logic. For example, things I'm considering doing is stuff like predicting location vectors for my arena opponents to determine the probability that they will run out of LOS within the next GCD to determine if I should take preventative actions. Another thing is that I would like to have a better spell/action system in place that better maximizes my selection of abilities to use. For example, I detect that the opposing warlock is casting a 2.0 second spell, since my GCD is 1.12 (after spell haste considerations) I decide that I can get a Dmg spell of before I need to cast the interrupt spell. I understand a world-state system "could" be very large dependent on what you monitor, but it could also be very simple. E.g. only track opponents and friendly position, facing, current casting, current trinket use, health, target, power, cooldowns. From those few states you could infer distance, reason action (i.e. if they both target you chances are they are going to burst you), monitor spell-casts and take preventative action.

[Apoc]

Or, you could use a BT and get the same effect with far better performance, and far less code. (Thus; easier to maintain and expand)

[!@^^@!]

Nitrogrlie, im just starting to getting into botting and behavior but to me it sounds like a behavior tree consisting of other behavior tree's would be the optimal solution for you as it support that there is more than one way to react to a condition with selectors and lots and lots of conditions.
If you haven't watched the movie on aigamedev i linked in the first post i would highly recommend it because it shows how behavior tree's are a middle ground between FSM, planners and scripts

-edit- beat to it by apoc, DAMN U! (btw now that i have you attention, is it possible to get a nick change? can't edit/change or even see my profile because of the chars)

[namreeb]

MMOwned - World of Warcraft Exploits, Hacks, Bots and Guides - View Profile: !@^^@!

[!@^^@!]

fine thats one way to go but the fine little profile button i have doesn't work at least :<

[nitrogrlie]

So, yes, my intent at one point was to use BT, but sadly my implementation made it become too rigid and it essentially became a FSM.

So on a related note, is anyone aware of publicly accessible code that provides a baseline BT library for C++? I scanned through the Game::AI++ source release and while it is applicable it is also very much over-the-top for a bot implementation. If no one knows of one I might write my own and potentially release it for everyone, but if one exists, why reinvent the wheel?

[!@^^@!]

I think Game:Ai++ is the only c++ code that is public atm and yes the code from Aigamedev is overkill for botting so i would recomend you to watch Behavior Trees for Next-Gen Game AI&mdash; AiGameDev.com (need a free signup) and just pause once a while and implement the features you think could be of use, like i did

[Apoc]

The AIGD BT implementation is actually incredibly well written. (It's written by one of the guys on the forefront of BTs)

It may be a bit 'over the top' to you now, but you'll realize just how nice it is to have a robust base to work from.

[!@^^@!]

Added links to a good codeproject article about marshaling