Jump to content
Lt Bull

Scenario designing with AI

Recommended Posts

5 hours ago, Kuderian said:

I have been experimenting with the new V4 features. The face, area fire and retreat are all great addition, I think.

Three AI questions:

Often with triggers you can achieve the same thing with both terrain and order triggers. Which is best in which situation?

How do you stop your Platoon HQ leading from the front in an advance move?

How can you achieve recon in attack/defence scenarios with the AI?

Thanks in advance.

Simple answer is terrain triggers are reliable in that the enemy triggers a response whilst unit order triggers are more prone to upset due to enemy action i.e. unit A gets KOd and therefore does not move. Depending on how you set-up the trigger your responding unit may or may not ever move subsequently.

In RL you can mix and match but if you defo want an event to occur and it is within combat range then terrain triggers are generally a safer bet. That being said it does depend on the size of the AI unit e.g. a AI group consisting of five tanks is more robust than an AI unit consisting of one tank. In that case order triggers, even within enemy contact still have a good chance of working, even of the AI unit takes casualties (although then if they take too many casualties then the unit may lose morale and stall). There is a good section in the manual which discusses the relative pros and cons.

It will all eventually come down to play testing as well which approach is appropriate for any given situation. Oh whatever you do, for your AI plan - keep it simple, like really simple. test it in scenario author mode then slowly start to develop its complexity.

Share this post


Link to post
Share on other sites
On 9/6/2018 at 4:15 PM, Kuderian said:

I have been experimenting with the new V4 features. The face, area fire and retreat are all great addition, I think.

Three AI questions:

Often with triggers you can achieve the same thing with both terrain and order triggers. Which is best in which situation?

How do you stop your Platoon HQ leading from the front in an advance move?

How can you achieve recon in attack/defence scenarios with the AI?

Thanks in advance.

 

what GeorgeMC said above.

It´s been some time since I meddled with AI plans, but generally I do not really see a rule of thumb for when using terrain or unit triggered orders. IIRC I mainly used terrain triggers to deal with enemy unit movements/actions (reaction to) and kept unit order triggers mainly for coordination within larger formations that are to move sooner or later. IE. to make sure that support units (triggered) don´t move forward (to new supporting positions), before the to be supported units ALL have reached a particular position and start (trigger) their next movements. Really depends on particular situation and intent.

Can´t remember that I noticed Plt HQ units (infantry or tank btw.?) leading in front of the actual Plt during advance in V4. Can you elaborate? Is the Plt. HQ and Plt units in their own/seperate AI orders group, or part of a larger Coy AI group?

With regard to AI recon techniques I usually set these up the same as human players normally do. Have small units (Bn, Coy and certain Plt.HQ or FO) with binoculars (having more than 1 Binoc prefered) overwatch from strategic locations (hills, high buildings with very good field of view) and keep them in the communications net. Counts for a an attacking and defending AI alike. Make sure they don´t shoot & fight unless in an emergency, so they only watch and report. For this to work well, I usually put them in their own orders group und move them only, when necessary. That could either be when threatened by the enemy coming too close (terrain trigger), or when during an attack a more forward overwatch position is secured by friendlies (unit order trigger). For more active recon techniques there´s use for the new area fire AI order (hoping a nearby yet non discovered enemy starts shooting) and everything that you would do as human player. If terrain allows, have small moving recon detachments (2-4 men) overwatched by both Binoc equipped and support units (MG, Mortars, FO), so they can help if the recon guys are engaged by newly discovered enemies. How to set all this up in an AI plan is a matter of repeated testing and very time consuming. Though keeping it simple isn´t much of an option here anymore and it all works best if you´re not having too large a force and have enough orders groups left, to deal with more detailed coordination stuff.

Share this post


Link to post
Share on other sites

Thanks @George MC and @RockinHarry for the informative answers.

The HQ that who was running ahead was an Inf. Pltn. HQ of an American Inf. Coy. The  whole Pltn. shared the same AI group. I solved the problem by first manually splitting the squads, secondly increasing the time to execute the order and finally painting a more triangular shaped order.

When painting orders, do you paint 'broad' strokes and let the AI decide on the best placement or paint small exact action spots?

And will the AI ever split and reform squads itself?

 

 

Edited by Kuderian
More

Share this post


Link to post
Share on other sites
39 minutes ago, Kuderian said:

Thanks @George MC and @RockinHarry for the informative answers.

The HQ that who was running ahead was an Inf. Pltn. HQ of an American Inf. Coy. The  whole Pltn. shared the same AI group. I solved the problem by first manually splitting the squads, secondly increasing the time to execute the order and finally painting a more triangular shaped order.

When painting orders, do you paint 'broad' strokes and let the AI decide on the best placement or paint small exact action spots?

And will the AI ever split and reform squads itself?

 

 

Thanks, I can compare the situatioon well with my scenario over at TSD3 as I have an AI US (mostly infantry) force set up to attack a fortified german player. http://www.thefewgoodmen.com/tsd3/combat-mission-battle-for-normandy/cm-battles-for-normandy/cmbn-v4-mg-vp-you-enter-germany-introduction/

IIRC I as well split up the US Plt HQ, since it´s actually part of full size squad A (using the disembarked mech Inf. Coy). I also put the small Coy support units (HMG and Mortars) away and attached them to the the Coy HQ in a seperate AI group. This also speeds up the Plt´s infantry squads beeing freed from their organic support for easier and faster movement between zones. Reforming of squads only occurs randomly (if HS´s/sections end in the same action spot coincidally). The AI isn´t capable of that and most likely rips formations apart more and more, the larger the AI group and zones. For Plt size zones, check page 1 of this thread where I elaborate on the "T" scheme idea.

Generally the AI only tries to move  - fill the forward edge of a zone and if broad enough, all AI groups units are placed forward. Otherwise if the zone also provides some depth, the excess units will be placed in 2nd row so to say. Personally I haven´t seen intelligent AI choices for filling and moving to the zones, so I try to limit size and width of zone as much as possible. I at least do that for Plt size AI groups if the overall force is not too big and with sufficient AI order groups. I´d likely give more freedom for an AI force in roughly Bn size, since there´s not enough AI groups to split it all up to Plt size groups. Moving a Coy size AI Inf. group from zone to zone at least requires as many painted AS in the destinations zone as there is single units in the group. No matter the final size and shape of zones, the AI will rip formations apart during movements and thus makes units most vulnerable when not in cover terrain, or with wrong movement orders.

Beside painting of zones also important is distance between zones and cover terrain in the area. IIRC leapfrogging starts if zones are apart 60-80m or so. To some degree you can steer AI unit movements by placing some shellholes (the smallest ones suffice), if a zone is not in actual cover terrain, or in the middle ground between zones. This is due to the unit magnet nature of shellholes. Walls and road ditches in destination zones need also beeing considered, as is small dips or rises in the terrain mesh (ditch locked AS, + or -) for beeing unit magnets. So if the zones don´t quite work during testing, also consider tweaking the map and not necessarily the zones alone. MOUT is quite a different beast and usually leads to disaster for an attacking AI, but while limited, there´s some solutions as well. 

Share this post


Link to post
Share on other sites
13 hours ago, RockinHarry said:

Beside painting of zones also important is distance between zones and cover terrain in the area. IIRC leapfrogging starts if zones are apart 60-80m or so. To some degree you can steer AI unit movements by placing some shellholes (the smallest ones suffice), if a zone is not in actual cover terrain, or in the middle ground between zones. This is due to the unit magnet nature of shellholes. Walls and road ditches in destination zones need also beeing considered, as is small dips or rises in the terrain mesh (ditch locked AS, + or -) for beeing unit magnets.

I haven't seen the AI react in any way to cover along the way when moving troops. Are you sure? To me, it seems the AI just gives random move orders in the general direction towards next move zone. It will happily run straight past shellholes, walls, hedges, etc. and then plop down on open ground.

Share this post


Link to post
Share on other sites
3 hours ago, Bulletpoint said:

I haven't seen the AI react in any way to cover along the way when moving troops. Are you sure? To me, it seems the AI just gives random move orders in the general direction towards next move zone. It will happily run straight past shellholes, walls, hedges, etc. and then plop down on open ground.

I meant shellholes & cover t errain generally at the leapfroggin range (if distance between 2 zones is larger than mentioned 60-80m) or within the destination zones. It´s the zones the Pixeltroopers stop anyway, either for beeing in wait mode to let the leapfroggers catch up and overtake, or final zone as said. Other factors left aside (getting pinned from fire ect.) Doesn´t work always off course, but I found this gives me a much more reliable and "safer" AI movement scheme. So when in doubt, I do not necessarily adapt AI order types and zones, I also give adapting the map some heavy consideration.

As you mention the AI move orders otherwise also look pretty random to me, but to some extend I could overcome parts of that by using specially shaped zones at intervals, as described on page one of the thread (intermediate T-Zone scheme).

 

Edited by RockinHarry

Share this post


Link to post
Share on other sites
1 hour ago, RockinHarry said:

 using specially shaped zones at intervals, as described on page one of the thread (intermediate T-Zone scheme).

+1  Interesting stuff.  Thanks for sharing. 

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×