Don Maddox
10 Apr 04, 11:46
AI is always a difficult aspect of a wargame to really get right. Programmers could spends years on end writing code and the AI could still be rather easily tricked once you understand what it's doing and why it's doing it. A while back, Jim Lunsford said that he would like to hear ideas on how it could be improved.
First of all, I've never been happy with the AI and I'm always looking for ways to improve it. Other than the features you mentioned, there are really no ways someone can edit the AI behaviors. However, the hard-coded behaviors are designed to mimic a modern commander's decision-making process. The "AI Commander" goes through IPB (Intelligence Preparation of the Battlefield), course of action development and analysis, and follows a targeting methodology for developing a plan to use fires. The tools you mentioned provide the "AI Commander" his mission focus. The "AI Commander" will sometimes suffer pangs of doubt and this will generate hesitancy on "his" part. He is randomly generated with a particular personality style that influences his decision-making style.
Again, I would love to improve upon it; so if you have any ideas please let me know.
Here is one suggestion that I have seen work quite well in other wargames.As I understand it, right now individual units in DA can be keyed to final waypoints and scenario authors can also set objective markers. One way to make things a bit more dynamic -- although it does make scenario authors work harder -- is to tie-in AI behavior with an event editor/trigger system. The way this is done is to have multiple AI "tracks." In each track, individual units can have a different final waypoint set for them (or they could use the same one depending on how the scenario author wants this particular unit to act). Activation of these tracks is tied to either a random percentage probability, or to a specific event that is triggered when a pre-defined occurrence happens during gameplay.
Example: REDFOR is basically programmed to advance to locations near the NE portion of Kansas City (this is track 1). During gameplay, BLUEFOR seizes a critical terrain location at another point on the map (we'll call this OBJ #1). An event has been pre-defined by the scenario author that gives the REDFOR a 60% probability of switching to AI track 2 if OBJ #1 is seized by BLUEFOR. In track 2, many of REDFOR's units (but not all) have different final waypoints, thus causing the REDFOR commander to switch part of his main emphasis.
This is really just a simple example of how AI behavior can be tied in with a simple event/trigger system to make a given scenario far more realistic. This isn't really "making the AI smarter" so to speak, rather the scenario author is doing some of the thinking in advance for the AI. The beauty of this system is that it needn't add an unwanted layer of complicated triggers to scenario creation. If the scenario author doesn't wish to use it, then each unit has only a single "track" defined and no events need to be input. Such a scenario would play out exactly the way it would right now. Allowing scenario authors to tie-in the final waypoint objectives for each unit to a specific track opens up a whole new world of possibilities for complex AI behavior. Devious scenario authors that anticipate certain strategies by the enemy can set up the AI to react in ways simply impossible right now.
The TOAW system uses a very complex event editor/trigger system that can do what I just described plus a lot more complicated stuff. DA is more limited in both scale and time frame, so such a system isn't needed. However, a simplistic AI track/trigger system could provide scenario authors with a very useful tool that could make the DA AI far more formidable in a wide variety of situations. With such a system the AI is limited not by the developer's ability to program it and make it smart, but only by the scenario author's devious ideas...:devious:
First of all, I've never been happy with the AI and I'm always looking for ways to improve it. Other than the features you mentioned, there are really no ways someone can edit the AI behaviors. However, the hard-coded behaviors are designed to mimic a modern commander's decision-making process. The "AI Commander" goes through IPB (Intelligence Preparation of the Battlefield), course of action development and analysis, and follows a targeting methodology for developing a plan to use fires. The tools you mentioned provide the "AI Commander" his mission focus. The "AI Commander" will sometimes suffer pangs of doubt and this will generate hesitancy on "his" part. He is randomly generated with a particular personality style that influences his decision-making style.
Again, I would love to improve upon it; so if you have any ideas please let me know.
Here is one suggestion that I have seen work quite well in other wargames.As I understand it, right now individual units in DA can be keyed to final waypoints and scenario authors can also set objective markers. One way to make things a bit more dynamic -- although it does make scenario authors work harder -- is to tie-in AI behavior with an event editor/trigger system. The way this is done is to have multiple AI "tracks." In each track, individual units can have a different final waypoint set for them (or they could use the same one depending on how the scenario author wants this particular unit to act). Activation of these tracks is tied to either a random percentage probability, or to a specific event that is triggered when a pre-defined occurrence happens during gameplay.
Example: REDFOR is basically programmed to advance to locations near the NE portion of Kansas City (this is track 1). During gameplay, BLUEFOR seizes a critical terrain location at another point on the map (we'll call this OBJ #1). An event has been pre-defined by the scenario author that gives the REDFOR a 60% probability of switching to AI track 2 if OBJ #1 is seized by BLUEFOR. In track 2, many of REDFOR's units (but not all) have different final waypoints, thus causing the REDFOR commander to switch part of his main emphasis.
This is really just a simple example of how AI behavior can be tied in with a simple event/trigger system to make a given scenario far more realistic. This isn't really "making the AI smarter" so to speak, rather the scenario author is doing some of the thinking in advance for the AI. The beauty of this system is that it needn't add an unwanted layer of complicated triggers to scenario creation. If the scenario author doesn't wish to use it, then each unit has only a single "track" defined and no events need to be input. Such a scenario would play out exactly the way it would right now. Allowing scenario authors to tie-in the final waypoint objectives for each unit to a specific track opens up a whole new world of possibilities for complex AI behavior. Devious scenario authors that anticipate certain strategies by the enemy can set up the AI to react in ways simply impossible right now.
The TOAW system uses a very complex event editor/trigger system that can do what I just described plus a lot more complicated stuff. DA is more limited in both scale and time frame, so such a system isn't needed. However, a simplistic AI track/trigger system could provide scenario authors with a very useful tool that could make the DA AI far more formidable in a wide variety of situations. With such a system the AI is limited not by the developer's ability to program it and make it smart, but only by the scenario author's devious ideas...:devious: