Well like many, scripting is basically my weakest point and now, having been at it for over 4 years, graphics, my strongest point, is suffering so I need to advance my learning curve and get a few things working at a basic level for me.

AI is critical! period.

Nodeless AI can give lots of good looking behavior routines from obstacle avoidance to hieght adjusting for both ground and air based AI and we can always get a bot to zero in on player thru vec_dist or scan functions then get it to 'hunt' down the player or avoid as needed but most require a lot of trials before the routine is found so all in all, the real problem with AI in GStudio is using routines to work with model terrains and levels (where the default block geometry provides both walls and content to work with) but models only can be accessed thru trace, dist and texture names I have found.

Nodeless AI is good to use in water pools, large rooms or confined outdoor areas where pathfinding is not an issue and adds a great deal to the believability of the game.

So what I am proposing here is a listed methadology from experienced users who know how to handle these issues and provide a layout of the methods employed to overcome common AI nodeless issues in both block and entity levels either as working or theory discussions.

To hunt down the player you can create "nodes" o the fly,
that store past positions of the player. These nodes can be used for an AI to trace and follow the
payers path, when he tries to escape.

You dont always need to implement a pathfinding, if you keep the levels and abilities of the AI (regarding pathfinding) simple.

But simply: the more effort you put into AI, the more interesting the enemies will get.

Raiden has offered avoidance code that may be helpful in this thread:

web page

Yes, Raiden posted a beauty but I havnt had the time to work with it though.

Here is a link to an old project I havnt updated or revisited but adequatly describes the node less AI I have derived or edited from Georges AUM series and other contributors. (7 mgbs)

It doesnt work as well published as it does in dev mode.

The test also incorporates Leonardo's bone anims and dust too I think as well as a number of other scripts from various users.

You will find yourself probably jiggaling around, just use wasd to move, w + r to speed up, mouse left to fire cannon, mouse middle to fire homer and observe the behavior of both the air and ground based AI from both a passive and active perspective.
Use the 'radar' to locate prey when closer, drive staright ahead almost to find them.
Firing the cannon will help to orient you to the AI direction.

All scripts are available.

The AI is from a direct contribution by a contributor whom I have forgotten but is either tuscharten, turrican or Domacales, I will get it right eventually and give proper credits as soon as I am able to.

Hope this helps to extend the discussion.

Anyone tested out the above linked demo?
Need more thoughts on the subject if theres any to be found.

Thanks for the link but I am too new to comment deeply.
From what I saw it all shows rather believable AI but the flyers are too locked into circling around, how would you work it so that AI only cirlces 2 times at most?

The circling around can be tweaked by editing its pan and period to hang around player untill a shot opens up.
Its all litle steps as shown in AUM series node-less AI, which need to be tweaked and tested untill it work rights.

Have removed the link as lots of downloads are happening but no replies (bad for my server bandwidth) so if any want to see it, I'll PM the new link.

Link back up now

Well, you could always help me with my new AI project. Using Fixed Effectiveness Projection & conditional probability. It's a longshot, but eventually will turn out well...

but aside from that, you could always use a very basic AI:

Code:

---

if (player visible)
{
    if (player close)
    {
        attack;
    }
    else
    {
        run towards player;
         - avoid obstacles;
    }
}
else
{
    if (I am where player was last seen)
    {
        if (player visible)
        {
            continue chasing
        }
        else
        {
            error("[censored]...");
        }
    }
    else
    {
        .go to where I last saw him/her
    }
}

---

I found this code to work generally well, along with some simple obstacle avoidance (such as Raiden has provided)