Combat Mission’s Tactical AI

As you may know, the computer opponent in Combat Mission is a combination of scripted (i.e. set up by the scenario author) high level behaviour and a non-scripted artificial-intelligence-driven low level behaviour. To give you an example, with the scenario editor an author can instruct a particular computer opponent group to go to certain places. These are called "orders" in the scenario editor. But the scenario author can't specify how the exact path will be or how the computer opponent group will behave during that movement.

A series of modifiers can be given to the orders for computer opponent groups, like emphasis on fire and/or movement. But down to "men on the map" the exact behaviour of the computer opponent is driven by the artificial intelligence. For the purposes of this blog entry, I will call that portion of the artificial intelligence in Combat Mission "Tactical AI" or "TacAI".

A notional graph for the emphasis in fire or movement for different orders is shown below. Please keep in mind the word "notional". The graph has no measurable units that we can't account for but just an index. It just reflects a trend in the compromise between moving fast with little emphasis on putting rounds on the enemy and moving slowly and getting as much firepower onto the enemy.


A more detailed description of the orders system in the Combat Mission editor can be found in the game's manual.

The purpose of this blog entry is to examine Combat Mission's TacAI and its ability to find out cover and concealment, and if that ability includes figuring out a path towards their destination.

So I designed a testbed scenario in which I am confronting a squad of German infantry against a platoon (-) of Syrian counter parts. The scenario's map is shown below.

The Syrian troops have to move towards the opposite side of the map, which features a ridge with German troops.

The lighter grass area in the forefront is a small ridge so to keep the Syrians out of sight from the Germans during the first seconds of the scenario. In the middle of the map, there is a ditch (midfield ditch) and to the right and left there are a forest and a built up area respectively. I ran this scenario multiple times, using the "scenario author" mode, which lifts the fog of war and allows the scenario author to see what the computer opponent is doing.

The Germans in their initial positions. Ready for battle.

Please keep in mind that this scenario is just a test to see what the TacAI does. There is no tactical lesson or anything else but an evaluation of what the TacAI will do.

The Syrians will win this encounter. The only question I care is how.

When the Syrians automatically deploy in a narrow sector (designed with the editor, narrow deployment and narrow objective) and have to attack a narrow sector on the opposite side of the map, there are two major options. One is to move through the center, exposing themselves in the middle of the map, without any type of cover. The second is try to divert towards the left and right of the map, to keep themselves out of sight and with a lot of cover.

Max assault orders, narrow sector of deployment and attack, at the pre-contact time. The red lines indicate waypoints.

Two main things can be pointed out from the previous screenshot. The TacAI, even before knowing where the enemy is (pre-contact) was making some sort of terrain reasoning beyond the generation of waypoints. In the screenshot above, you can see how the squad on the left was directed towards the ditch (which offers reasonable cover). Also and most surprisingly, the TacAI directed the squad on the right to move out of the "sector" and use the wooded area to the right.

Max assault orders, narrow sector of deployment and attack, post-contact time. The lines indicate waypoints.

Post-contact, the Syrians react to being exposed in the open and a survival type of behavior kicks in. In this case is to reduce exposure by crawling towards the destination. These type of reactions are built-in even for friendly troops.

Now let's go to the opposite type of order, "dash", and see what happens.

Dash orders, narrow sector of deployment and attack, post-contact time. The lines indicate waypoints.

As seen in the screenshot above, this time around the TacAI had no regard for finding cover and just moved at high speed towards the destination. Pretty much what the manual describes for this type of orders.

The middle of the spectrum of orders is the "advance" one. It is supposed to be a compromise between fire and movement.

Advance orders, narrow sector of deployment and attack, post-contact time. The lines indicate waypoints.

The advance order gave the TacAI an emphasis on the use of the midfield ditch. In any of the runs of this scenario I observed the TacAI to move out of the narrow sector.

I also ran this scenario with the assault and quick orders and I observed intermediate behaviors between the ones described above.

So far, the main conclusion is that given the right orders the TacAI can plan to use covered positions on its own and even moderately venture out of it's sector.

I was not expecting the TacAI to thoroughly use the wooded or built-up areas (i.e. out of its sector), and although that would be awesome, I understand the design's philosophy. Who wants to end up with a computer opponent that hugs terrain very far away from its objective or deployment area?

But nonetheless, I ran the scenario a few times using a wide sector (Syrian deployment and objective areas covering the entire width of the map). The results of one of the runs is shown below.

Advance orders, wide sector of deployment and attack, at the pre-contact time. This is just to show the spread of the Syrian forces at the beginning of the scenario.


Advance orders, wide sector of deployment and attack, pre-contact . The lines indicate waypoints.

In the screenshot above, the squads are moving directly into the built-up area. A have the suspicion that this more due to the fact they were deployed just in front of this built-up area than anything else. Maybe also the foot path offers some additional incentive. But the waypoints end up in the village. I was surprised to see that the TacAI chose to use just the edges of the wooded area.

Advance orders, wide sector of deployment and attack, post-contact time. The lines indicate waypoints.

Either woods are not so cool anymore, or they pose a significant obstacle to movement.



Leave a Reply

Your email address will not be published. Required fields are marked *