AI

[LiveNoMore]

I'm trying to learn the ins and outs of using the AI for scenario building. Can't seem to keep the defenders from leaving their positions and heading towards the attackers. Does anyone know of a better tutorial then the few pages that came with the game?

Or even better, how do I keep them in their assigned positions?

Thanks Guys

[George MC]

Check your AI plan - see the orders for the group? Check you don't have a spurious painted action square or squares. No painted squares on the map means your AI units will not move - painted squares associated with an AI group = movement.

Other top tip - start your scenario small and create broad brush AI Plans. DON'T get involved just now with trying to get carefully coordinated AI plans with units etc moving in perfect harmony. Down that road lays madness and frustration. KISS principle.

This MANUAL may help as may this small test scenario which will give you an idea of how to create an AI plan. Just open it in the editor and check out the German side to see how I've done the AI plans.

Good luck

[LiveNoMore]

Thanks for the help! The manual looks good.

[George MC]

Nae bother - if you've any questions feel free to ask away. There's more than a few designers on here who'll be more than happy to give you a helping hand. Plus it's always useful to pinch ideas...err learn from others what they do ;

[Vinnart]

George, First thanks for making that manual. I read it, and it was helpful. I am trying my hand at AI planning for the first time to make my “Death’s Head” map playable as a single player ME QB. The map has many different victory point objectives spread out with lots of possible ways to approach it.

is it really better to paint just over objectives, and let the AI decide best way to get there, or to paint in paths of advance to get them there to make sure they stay in good cover? I really want the AI to make good choices to advance smart. The map has many different ways it can be taken either along a wide front, or in flank attacks.

You say you recommend wider painted orders, as opposed to lines. Why? When I look at the AI plans from LJF’s maps he seems to use more painted lines.

What setting is best to get the AI to keep some units back on an objective to defend while some of the group advances, or does a group need to be designated to defend a rear objective to get them to keep some units on a flag?

What is major difference between advance, and assault as far as AI is concerned?

I see the AI will divide a QB force into groups on it’s own, but in my last test group 1 was missing. I had another test where AI groups were missing too. Any ideas why? Is this normal? Right now I have the AI force in 6 groups to get them to spread out across a broad front at order 2. From there I want each group to move in different paths toward the objectives on the enemy side. I also need them to keep some units on their rear objective to occupy.

Any other particular tips regarding meeting engagement battle planning I would appreciate. Thanks.

[Paper Tiger]

Vinnart

if you can hold on for a little while, I'll be putting together a series of posts talking about AI planning in great detail which will cover all those points. It'll start up soon after JonS has finished his Sheriff of Oosterbeek thread.

[Vinnart]

Paper Tiger from playing around with it this evening I have been able to answer most of my own questions. I do look forward to reading the AI tutorial that you will be doing. The only one I havn't figured is best setting to get them to keep some units on rear objectives while some of the group advances. The work around is to just make a group to move to rear objectives, and stay there. I actually got the AI putting up a better fight than most ME maps i have played. The key I think is in making more groups with clearer moves via more orders. From my expereiments I think more order waypoints to keep them on better paths to objective is beter than leaving it to the AI.

Here is the plan 1 which is a wide front approach. All group set-ups are in blocks where the icons I am showing. I would rather use more of a line for set up as that is how I would do it as a human. The set up is set to active. Order 2 is set to dash, active, and unload passengers. Orders 3 and 4 are a mix of advance, and assault set to active. The map is basically a large long hill that has different elevations, and ridges on it. There is lots or rocky open ground with patches, and lines of trees. Plan 2 will be a flank blitz taking it from left to the right, and plan 3 is the same from the opposite direction. Plan 4 will be a double envelopment. In al scenarios I want the AI to take, hold, and be ready to defend objectives blue, and green from this side.

I like playing it in the senario author test mode. It is like watching another person play in real time like I envision being able to observe multi-multiplayer RT.

[Paper Tiger]

Paper Tiger from playing around with it this evening I have been able to answer most of my own questions. I do look forward to reading the AI tutorial that you will be doing.

Well, to be honest, there's probably not going to be much in it that any determined player can discover for themselves after spending a bit of time watching the AI do its thing in scenario author mode. Glad to hear that you've found the answers to your questions on your own.

[Vinnart]

PT when you do your tutorial it would be helpful I think if you include info about making QB ME. I am finding it is a bit different from scenario making in regard to groups, and their limitations. The biggest difference, and challenge I am finding is that in a QB ME the amount of units will be different each time where as in a scenario they are constant. Because of this the flexibility for creating is easier for scenario than what I am doing.

The problem I am coming across is the AI not using all the groups unless there are more units. I have made 6 groups in the set up shown in the graphic that work in attacking good if the setting is set to large. Here all groups follow their plotted paths, and orders. If the setting is set to medium then only groups 2 – 6 will show up, with group 1 absent. If the setting is set to small groups 1 and 6 are present, and 2,3,4,5 are absent. This obviously creates a problem, as the plan cannot be fulfilled all the way unless all the groups are there. In scenario design it is constant so they will all be there all the time. I am also assuming that to have the AI keep some units on back flags a defense group must be created, as it does not appear defense, and attack can be in the same group. Part of the challenge is in the amount of VL’s in the map. I don’t think I have ever seen a ME battle with more than 3, or 4 VL’s. The fewer VL’s the easier. Most will have one VL in rear; so one group can be set to defend, while one or two groups can attack. Recently I made a variation of the map that has eliminates all the center VLs leaving only four (blue, green, red, and orange). This should be easier in regard to the amount of groups needed. From what I have seen though breaking the force down into more groups gives a better challenge to the player, but groups are limited for QB ME.

In light of the way groups work for creating ME battles I think an improvement can be made in the computer separating the force into the groups no matter the amount of units. Even if it put a single unit into a group it would be better than it eliminating the group altogether.

I will have to rethink my groups to make them less it appears. One group can be set to cautious, and be programmed to bring up the rear, and sit on flags to occupy for points. As far as the attack goes the minimum I think I can keep the attack going good is four groups. I will have to see if it attacks as good as I have it set now, which is more precise following the best terrain. Due to the size of the map if I can get it to fit into five groups it should accommodate a medium size force better as 5 groups showed with that setting. Med. Size force is the minimum one would want due to the size of the map. Currently as I said I only get the six groups if it is set to large.

When it comes to AI planning in general I think it is rather straight forward, and user friendly. It seems not much different than potting orders, and waypoints when playing, but less precise than what can be done as the player. Understanding the subtitles of the settings I look forward to hearing more about.

[Paper Tiger]

Vinnart

I'll include what I know about AI for QBs. I've built up a private collection of QB maps, ME included, and have evolved my own systems.

[Vinnart]

Paper Tiger, that would very helpful. In regard to the particular problem I have encountered with the computer excluding groups I think I will just have to play around to see what is the max groups I can use to best effect without losing any for medium size force. I’m going to knock it down to five, which should work for at least a medium size force. I say that because I was getting up to five groups with always group 1 absent. One group can be defense. I will try plotting order 2 for defense group to objectives blue, and green to dismount set to cautious. Also, I see that you can paint an order over an order. If that works then I can paint order 3 over all objectives. This should leave some units on blue, and green while other units in the group move toward each objective slowly. Is that the secret to having a single group leave some units while others move on? If so, then that improves the flexibility of each group. I have been testing with mech infantry, so I also need to find out how groups are divided with an armor exclusive force.

Paper Tiger, that would very helpful. In regard to the particular problem I have encountered with the computer excluding groups I think I will just have to play around to see what is the max groups I can use to best effect without losing any for medium size force. I’m going to knock it down to five, which should work for at least a medium size force. I say that because I was getting up to five groups with always group 1 absent. One group can be defense. I will try plotting order 2 for defense group to objectives blue, and green to dismount set to cautious. Also, I see that you can paint an order over an order. If that works then I can paint order 3 over all objectives. This should leave some units on blue, and green while other units in the group move toward each objective slowly. Is that the secret to having a single group leave some units while others move on? If so, then that improves the flexibility of each group. I have been testing with mech infantry, so I also need to find out how groups are divided with an armor exclusive force.

If there are any exceptional ME QB with AI more than 2 or 3 groups I would love to check it out.

[Paper Tiger]

If something is worth saying, then it's worth saying it twice

[Vinnart]

Note to self. Post more at 2am with tired mind, and eyes. It improves proof reading skills . I write in word, and paste doing it twice without even realizing. The sad thing is I can’t edit it.

In regard to using more groups more complications. With armor only med force it is only using two to three groups instead of the five I went down to.

Next, I will try keeping the plan with five groups, and make alternant plans with four, three, and two groups with the hope that the computer will select the plan that best matches the amount of groups it only wants to utilize. Otherwise it misses parts of the plan when programmed for more groups.

I was right to assume that stacking orders over objectives will leave some units on VL while others move on to next VL. At least I know now I do not need to designate a group strictly for defense.