A disappointment of sorts

[SteveP]

One of the things that I did not like to see in CM1: when the AI was on the attack, it wasn't unusual to see the assault being led by HQs, or specialist teams, or the like. Who immediately got killed, thereby giving the AI something of a handicap for the rest of the scenario. It's also ugly to see, though playing at a high difficulty level does spare one some knowledge of this until you reconstruct things at the end.

I had hoped that this problem was solved in CM2, but apparently not. Perhaps it's just not on the list.

[Ranger33]

I agree, it seems the "AI" basically consists of going from point A to point B most of the time. A good map designer can overcome this to some extent in a scenario, but it really shows in the QB's.

[MikeyD]

I thiink the AI is coded to keep some units types behind the other but as the old saying goes **** happens. The guys up front get surpressed or have to move around an obstacle and the guys behind catch up to 'em.

[BigJ62]

Mostly what I have been seeing is that HQs do stay behind the squads. But when there is not much else to send forward then the HQs take the lead.

[Peregrine]

No matter what happens the AI will never be good on the attack unless it has a significant force advantage and the scenario is designed well.

When the time comes that we have a true AI that can credibly attack we will be to busy fighting Terminator style robots to play computer games anyway.

[LongLeftFlank]

Yes, in CMSF, the HQs and MG teams were DEFINITELY the last to move in each sequence. But because the destination for each unit for each AI movement command is randomly selected from among the various destination squares, they might well seem to be ending up at the "tip of the spear" a lot of the time.

Which in WWII isn't as unrealistic as it sounds. C2 wasn't as big at platoon level as it is today, and subalterns on both sides were expected to "lead from the front", or at least had the idea that they did. The old cavalier tradition hadn't entirely been beaten out of armies by the WWI carnage, even though the sabres and horses had long gone by the wayside. Hence the disproportionately high casualties among lieutenants and captains.

[MarkEzra]

Mostly what I have been seeing is that HQs do stay behind the squads. But when there is not much else to send forward then the HQs take the lead.

Yes. I watched more than a few AI attacks in "author mode" and say with confidence that hated CMx1 HQ charge has indeed been addressed. But Like MikeyD BigJ62 says... Shxtt happens

[SteveP]

OK. I will withdraw the charge pending further review. And really it was a small complaint. Overall, I am impressed with the quality of the AI, with good scenario/QB design.

[Battlefront.com]

We've gone through progressive tweaks on this behavior, one of which happened during the final weeks of testing. It's really easy to see (or not see) such behavior when using Scenario Author Mode because you watch the AI's forces with zero Fog of War. Your units still have FoW, but as the player you don't. This is a fantastic feature for debugging just these sorts of things.

Note that there is another way that the AI can be tripped up when it comes to HQs and what not. If HQs are exclusively assigned to an AI Group, and the AI Plan incorrectly has them ordered forward ahead of the AI Groups with infantry, then for sure HQs will lead the way. That's because the AI can only control behavior within an AI Group, not behavior between AI Groups.

Steve

[SteveP]

We've gone through progressive tweaks on this behavior, one of which happened during the final weeks of testing. It's really easy to see (or not see) such behavior when using Scenario Author Mode because you watch the AI's forces with zero Fog of War. Your units still have FoW, but as the player you don't. This is a fantastic feature for debugging just these sorts of things.

Note that there is another way that the AI can be tripped up when it comes to HQs and what not. If HQs are exclusively assigned to an AI Group, and the AI Plan incorrectly has them ordered forward ahead of the AI Groups with infantry, then for sure HQs will lead the way. That's because the AI can only control behavior within an AI Group, not behavior between AI Groups.

So, I will also withdraw my speculation that this was not on the list for improvements. It's certainly possible that my reactions reflect my particular distaste for this behavior (like half the posts on this forum ). With CM1, I really thought it weakened the AI. In CMBN, I'm not so sure about this. It may be that it looks wrong more than anything else. But it isn't just a problem with scenario design or something like that. I just don't think there is yet any governing rule that keeps HQs from dashing off to the next Map Zone even if they are the only ones going there.

[Battlefront.com]

As I said, there is code in there now (which can easily be seen using Scenario Author Mode) which shows that HQs and support weapons are held back while fighting units move forward. This is definite in the game now, so there's nothing for us to add since it's already in and working as it should be (within reason, of course).

What the AI currently lacks is an ability to coordinate forces between different AI Groups. If the scenario designer groups things in a particular way, or doesn't "debug" the script well enough, then bad behavior is quite possible. This is, however, not the fault of the AI because in this case it's following the designer's orders.

In the future we will give the scenario designers more tools to control behavior and encourage coordination. That will improve the AI's performance, especially on the attack, in many ways.

Steve

[Sgt Schultz]

Thanks for the info Steve.

I have noticed a HUGE improvement over CM1 during my endless scenario author tests.

If you set up the plan right, the HQs stay well back up to the point of real contact. I aborted my first few test runs because I was sure my Co/Btn HQs were left out of my orders group. Why? Because they did not move out until up to 1:00 after the rest of my group.

I have seen Plt HQs allow the lead elements to just barely get out of C2 before moving out.

The scenario designer has to hold the AI's hand quite a bit during any approach marches to ensure proper interval. From the very small sampling I took of existing scenarios, many plans are very simple. This may lead to HQs taking the lead when the lead elements run into fire where the designer did not anticipate.

BFC gave us a lot of orders to use with each of our piddlin' eight groups.

Use 'em.

If I could ask for any changes, it would be for two more groups.

-

[LongLeftFlank]

Two more? I'd want at least 20 for a two company sized action!

I'd also like to see a slight bias of units toward destination squares that are nearest their current Locations. In other words, if Order 1 has my AI Group move to phase line Red and Order 2 has them move to phase line Blue, I don't want them zigzagging across the entire line of fire because their destination was randomly selected.

[BigJ62]

What LLF said, please put some logic in to go to nearest objective painted so I can cut down on number of groups. Also double number of groups and orders would also help with complex AI attacks.

[SteveP]

Lots of good feedback here. I know what I am seeing. But I also know that I have a lot to learn about AI scripting, and that this is one of the most interesting new features of the game. So I will go back to learning and experimenting with the scripting myself and see what happens with different scripts. It does seem to me that if no one has created a comprehensive guide to scenario creation yet, there will have to be one at some point. Perhaps too many people are re-inventing the wheel.

Most of all, I am pleased to learn that this is indeed something BFC has worked on.

[mjkerner]

In the future we will give the scenario designers more tools to control behavior and encourage coordination. That will improve the AI's performance, especially on the attack, in many ways.

Steve

I've been (re)reading up on AI planning, and the intricacies of working with movement orders and "stance". This comment of Steve's caught my eye. Can't wait! I hope it's coming with CW.

[Sakai007]

I will third the request for more groups. As I delve farther down the rabbit hole that is scenario design, I am finding AI plans to be the most fun, and most challenging, aspect of designing missions. You can really frustrate a player with too good of an AI plan, and this is not the objective of the mission, fun is. Unless it's a hopeless defense type scenario, the player should win right??? Making this happen sometimes means I must program 'mistakes' into the AI plans that, should the player see and exploit these, makes for a fun and victorious scenario.

I find CMSF to be GREAT training for building in CMBN. The range and lethal nature of weapons like the TOW and AT14 make it critical to time and coordinate AI planning. In CMBN, ranges are shorter and engagements seem to last a bit longer, meaning a little less precision is OK.

[StoneAge]

I totally agree with LLF 20 groups is more in the ball park.

Just think of when you are attacking would you move a full platoon and attached weapons teams across a field all in one go (more groups just gives better control).

The other things I want are:

1) Setting a delay for when a pre arranged barrage comes down.

2) Some triggers so I can have a platoon in reserve and when enemy crosses x squares start their plan.

i.e. paint two areas on map name them, create two plans for the reserve platoon, when enemy activate one run this plan and delete the other.