How far can the TacAI go?

[WMurray]

A general question, I was wondering to what degree the TacAI can represent general strategies for meeting engagements such as (1) find the enemy, (2) fix them, and (3) maneuver to engage them on the flank, envelop, etc. (Or correct me if me if that's the wrong strategy in the first place). Are there meeting engagement scenarios in the original ATF or the newer ATF games where friendly and enemy units both have a lot of room to maneuver?

I would think the only way you could handle these kind of scenarios is to allow for alternate enemy COAs, and then once you had determined one, then act upon it. But it seems like that would require sufficiently closed terrain to limit the number of alternate COAs to just a few. With LOTS of room to maneuver you couldn't anticipate them all and then this kind of meeting engagement scenario would be beyond the limits of the current TacAI, or any other game TacAI most likely.

My understanding of the ATF TacAI is that it represents COAs as phases. Each phase can have begin and end conditions. During a phase units follow paths, which may also require triggers to initiate movement, where the trigger is either a time or a boolean combination of other conditions (e.g., number of troops, presence of troops in a region, etc.). But then I haven't actually tried to build any scenarios, so please correct any misconceptions I have about the TacAI, here, too.

With the answers to these questions I might finally understand the TacAI. Not that it will help me, but I can admire the beast all the more.:)

Thanks!

[Pat Proctor]

There is one meeting engagement in ATF, Iraq 3: Salman Pak, I believe. Of course, if you make an error in building your defense, they all become meeting engagements, don't they? :) There may be one in TSATC. I just can't remember.

There is a little teminology mix-up here. Let me see if I can clear it up.

Enemy COA (AKA Battleplan): This is built by the scenario designer, and represents a single enemy "set" (The enemy units and hierarchies, the faction AI, enemy obstacles, and enemy mines). You can have any number of enemy COAs. The game pics one at random at the start of a scenario.
Faction AI (Script): For the military folks out there, this is most analogous to an enemy OPORD. It is actually structured kind of like the scheme of maneuver for an OPORD. The Fation AI has multiple phases.
Phase: Inside each phase is are decision points. Each decision point is triggered by a condition on the ground, and executes a series of orders. Some orders in a phase may never be triggered. Some may be triggered once. Some may be triggered numerous times. A phase also has a "Phase End Condition" or conditions. When all of these are met, the phase ends and the next phase begins.
Condition: Conditions are everything from time triggers to locations on the ground that the enemy watches to see what forces are there to statuses of the enemy's forces. Conditions trigger orders.
(Hierarchy) Orders: The faction AI can give any order the player can give to his forces. He can change a unit's formations, give a unit a path, tell a unit to stop or go, or give a unit a mission.

Based on all of that, here is how a movement to contact is built.

Enemy COA (AKA Battleplan): Each battleplan contains a different enemy set, with a different mix of recon, fixing, and enveloping forces. The Faction AI script might vary for each unit based on this force structure. The enemy is given a base path which just sends it straight to the other end of the map.
Faction AI (Script): In the faction AI script, a bunch of paths are stored. The paths are enveloping paths, that start at differnt points on the map, and aim toward other spots on the map. We will watch the entry points to those "aim points" for the player's forces, and trigger these paths based on what the enemy does.
Phase: We'll use your phases: (1) find the enemy, (2) fix them, and (3) maneuver to engage them on the flank, envelop, etc.
Conditions and (Hierarchy) Orders: In Phase I -- When we spot the player, the recon stops. Depending on what "Event Box" we spotted the player in, we select a path from our pre-planned paths, and assign a path to our fixing force to intercept the enemy. Phase ends when this order is given. Phase II -- Based on where we see the player main force, we select one of our planned paths and assign it to our flanking force. This phase ends when fixing force is in contact. Phase III -- We continue to watch the player's forces. If the player adjusts his main force, we adjust the path of our flanking force to envelop him. Phase ends with flanking force in contact. Phase IV goes on for ever. It is "complete the distruction". We continue to watch the player and our force levels. If we are in trouble, we comit our reserve by using one of our pre-planned paths to move it into position, or move helos to do a deep attack, or 100 other wicked things we can do to the player. Many of these orders will never get executed. It depends on what the player does.

This is an over-simplification. There is a lot of moving pieces missing, like how we position our artillery, etc. But this is exactly how a real force fights. We have a plan with phases and definitions of when the phase begins and ends. During the phase, we watch conditions to understand what our opponent is doing. That triggers our actions or decision points, causing us to direct our forces to do certain things.

That having been said, a meeting engagement is the most complicated thing to plan in ATF. An enemy attack deals with a static player. A enemy defense only deals with repositioning forces or moving reserves. A movement to contact is VERY fluid and tough to do well.

In general, the simpler the script the better. The more DPs you have, the slower the engine gets and the more like it is that the player will do something you didn't plan for, causing the enemy to look dumb.

[WMurray]

Very impressive. :surprise:

But, fortunately, I think I'm quite safe until you patch the friendly fire model in. :)

How is the combined arms part of a scenario handled in this case? E.g., do you select a corresponding decisive point for each engagement path from which the friendly forces attack the enemy once they are fixed, based on where they are found to be on the battlefield, and then based on that, use battlefield calculus to determine when the artillery should land to suppress or attrit the enemy? E.g., the friendly force has x meters to travel so I better plan the fire missions to be called in at a certain time so the artillery lands on the enemy at the same time as the main force arrives within direct-fire range of the enemy forces?

In general is the combined arms synchronization up to the scenario designer or does the TacAI provide some help there?

Thanks for your help understanding the TacAI.

[Pat Proctor]

The artillery is run completely by the AI. You can control the starting and stopping of artillery by giving and removing the "Do Not Emplace" order, but you cannot target artillery.

The game goes through a complex set of decisions in real time to decide what to shoot at, based on where the proponderance of enemy forces are, what the whole force sees, and the relative value of targets. It is a little ugly, but I would classify it as IJW ("it just works"). :)