Infantry Wheeling

I'm in the process of rewriting the infantry battalion code. One major section has to do with the battalion wheeling during a fight, (not while marching), to face a new or greater threat than the current one it faces. This behavior has almost always struck me as bizarre. For the most part, the existing code is a messy jumble. In situations like this, my first instinct is to take a hatchet to it, excise the malignant growths and write my own.

So the question is; under what circumstances would a battalion wheel or not wheel?
Would a unit ever wheel while it was under direct fire, even from arty?
If it was taking flanking fire, wouldn't the unit fall back rather than try to wheel to face it?
Would it ever wheel and present its flank to cavalry?
If it was faced with two advancing threats, one in front and one on a flank, would it wheel to face the greater threat or just fall back?

I can't think of any historic reports of units acting contrary to the above situations. Does anyone have any thoughts or information to share on this topic?

I noticed that the actual battle formation at times looked like a bunch of pick up stix at time. I am glad that you think you can fix it.

I do not know the parameters of the game but I would think that the reason to not fall back and wheel would be if they are supported or supporting other units.

What stance are they in would affect the way they would react/move under fire or to a target.

The individual units strength, morale and training.

I think a commander would look for the trade off before wheelin and I would doubt that they would expose themselves to danger when doing it unless they are facing an exhausted unit that was not making any hits or threats to them.

Where are they in the line might be a factor too.

Hopefully you understand what I am getting.

Can't wait to see this.

See you tomorrow.

All of the parameters you mentioned except for the unit's position in the line are readily accessible. Many of those could could be used to determine if a wheel would be successful if the enemy was near but not an imminent threat.

Interesting subject, one that I agree needs addressing.

First, can we completely remove the wheel about the flag action and make a unit always wheel on one of its flanks?

Can we remove any animations where sprites do a Michael Jackson and moonwalk backwards? Seeing this bugs the heck out of me.

I'd need to look up some drill manuals to see how a battalion formed in line could even hope to wheel backwards! I suspect a battalion would need to form a column of companies, about face and march to the rear. Under short range fire or the risk of a bayonet advance such a manoeuvre would be well-nigh impossible except to the best drilled and disciplined troops (say Exp6 and higher). A disorderly retreat would probably occur, if not the wholesale rout of the unit concerned.

A unit in line might be able to do an about face and march to the rear but again the effects of close range musketry or a bayonet advance into the exposed flank do not bear thinking about.

I think our main problem stems from a basic game design flaw where the game engine was intended to cope with ACW firing lines standing in fairly close terrain such as woods and thickets. In these circumstances a battalion wheeling to face a sudden local threat does not seem like too big a deal. But the AI is also programmed to attack in a very ACW-like manner, that is to always feel for the flanks. Even as players we do this when really flank attacks needed a lot of preparation and often required planning a day in advance and were conducted at least at the division level and more likely at the corps level (e.g. Soult at Albuera). An action where a force attempted a flank march without proper planning and co-ordination often failed such as Junot's flank attack at Vimeiro. A commander needs to have a good idea of the ground he plans to move a large body of troops across.

We simply have too many flank attacks in our games.

Can we do something about this issue first?

After that, I think a unit should never wheel once it is engaged. A wheel was just too cumbersome and slow a manoeuvre. You can look at the foot guards doing company wheels on the horse guards parade ground on Her Maj's birthday and this is with:

1) light uniforms and modern light rifles
2) on a perfectly flat clear hard ground
3) with troops who've practiced for the event weeks in advance
4) no one shooting at them

A company takes about 30 seconds to wheel, a battalion might take 5x that long.

We should slow down all our manoeuvres - maybe we need to lower the drill skill numbers to about 1 or 2 for everyone? 2 for elite troops, 1 for everyone else?

If what you a are talking about is a line of battalions in line firing and suddenly the middle one pivots a flank out 45deg just to get its sprites to face a target, then we ought to stop the game doing this.

Given that I think it would be extremely hard to do any kind of rearwards wheel under fire from a flank perhaps we should make every unit fired on from flank in musket range retire in disorder about 150 yds? This would encourage players to hold second lines in reserve, though what might we be able to do to help the AI think differently?

I think though the main issue is really a wider one. What we should be asking is not what did units do when flanked but - were units EVER flanked at all? About the only instance I can think of is the destruction of Colborne's brigade at Albuera, and here the flanked units did nothing and were hit in flank and destroyed.

I suspect we are asking the wrong questions to ask if battalions should wheel this way or that, perhaps whole brigades should advance or retire?

Its a tough set of questions but discussing it and also finding out what we can and can't do with the game would help.

These are all good points.

Flanking maneuvers in small (brigade-sized fights) were not uncommon. They were mostly rear guard actions where the attacking force would employ this technique when available. I can't think of any instance when the defending battalion didn't retreat. Division or corps sized flanking while not as common did occur. The first day at Eylau had the Russian rear guard fall back when Augereau's divisions began to arrive on the field and started to flank the Russian right. The second day saw Davout's corp perform a large scale flanking maneuver against the Russian left. In both cases, the Russian troops fell back rather than try to shift their position to face the new threat.

Flanking an enemy out of a position was a concept of the 18th century and therefore the ACW. One of Napoleon's innovative concepts was that winning geographic locations and forcing the enemy to retreat wasn't important, the destruction of the enemy army was the primary goal. That's why he delivered the killing blow in the center of the line whenever possible. You're right that players mostly think in 18th century terms instead of like Napoleon.

I can't imagine wheeling a battalion, regardless of its quality, while it is under direct fire or in close proximity to the enemy was ever contemplated by the officers. Rather they fell back and then reformed to face the new threat.

The AI's proclivity for trying to turn the flanks, is built into a part of the code we don't have. Although this behavior does fit the 1805-7 Austrian and Prussian forces, it doesn't fit the later period of warfare or ever reflect French tactics that were in general use. In any case we are stuck with it.

The wheeling behavior, with half the unit moonwalking and the speed of the maneuver is not something we can modify. We can, however, limit its use. We can program the AI to fall back or retreat rather than wheel. The brigade AI uses a double line formation as its fighting formation, so having a 2nd line of defense is not uncommon. Having a front line battalion fall back might not be too harmful to the AI. It's something we'd have to test though.

Having the entire brigade fall back is more tricky. Once a battalion is engaged in a firefight, it has a tendency not to disengage, or else it briefly disengages and then reengages. If none of the battalions are firing at the enemy, then making the entire brigade fall back is not a problem.

If we start by making infantry fall back when flanked that would be something to begin with.

Can you inhibit the wheels during a firefight where the AI tries to get its line to better face a target?

Can you inhibit the wheels during a firefight where the AI tries to get its line to better face a target?

I'm not sure if that particular behavior is part of the AI code that we have. The wheeling code we do have seems to have to do with turning towards another, more dangerous target, than the one it is currently facing.

I've worked on the dll and made some significant changes to the wheeling logic. Troops tend to move obliquely now and fall back when being flanked. I think it's at a point where we can test it tomorrow. Calling all guinea pigs!

I've been playing a lot of SP games over the last week or so and with the latest dll I have noticed infantry run to the rear much sooner than they used to, even in face to face firefights.

Could your latest change, Kevin, have accidentally fired up some broken code elsewhere? Units run away from ordinary firefights now as if they were under very heavy flanking fire, maybe even sooner.

I have done a couple of helicopter camera games and just watched the AI doing its thing to itself, and battles are very fluid with no discernable battle lines at all and units zooming all over the place. It's not encouraging.

I think you may be seeing the result of overlapping units falling back. I was worried that instead of just one of the units falling back, both would do so. Is that what you are observing? This is new code I added. The original checked for overlapping units, but did nothing about it. I'll try adding something so only one unit will fall back.

It could be that they were overlapping, yes. The problem seems to be that when a unit falls back from the overlap code it ends up overlapping another unit. And if a player-commanded force pushes on aggressively against these tangled units an enemy line can be shooed away quickly. Quite realistic if you are re-fighting Jackson's flank attack into the Union camp at the Wilderness but not very Napoleonic-feeling!

The AI still seems to park units overlapping each other. It does not seem to be aware this is a no-no.

Right now I think a player-commanded force has a big new advantage over an AI force because a player can TC battalions and park them adjacent and not overlap. When the AI tries to either flank a player or park several units close together to try to concentrate fire it falls foul of the overlap rule.

On an unrelated subject I am still seeing weird AI division level command decisions. I still see guns and cavalry brought forward and the infantry parked a mile or so away and held back until the cav and arty is destroyed, then the infantry attempt an attack unsupported. I have replays saved that show this in about 3 div vs div battles now. Its a worry.

The division strategies, (there are a lot of them), lump cavalry and artillery together, i.e. one being interchangeable with the other. Almost all the division layouts have them in the front center of the division. Obviously, the creator didn't have the slightest clue about 19th century warfare. This is actually a piece of the code non-programmers could change to see if different division layouts would yield more realistic behavior.

I'd be up for trying that if you could give me some guidelines and which files to edit and how. I presume its drills.csv.

No, it's one of the files used to create the dll. The particular piece of code is such that it is very easy to change where various brigades deploy. All that is required is a text editor. I'll email it to you.

Hi all.
Just a suggestion for Uncle Billy: don't waste your time, mate.
If you'll follow the current AI dll coding track, you'll just waste it, guaranteed.
Your progresses will be limited and your ambitions frustrated over and over again.
Those kinds of huge if/then/else behaviour trees going from "human controlled?", "has target?", "moving?"... up to the underwear color of battalions officers, are always unmanageable.
One might be able to catch warnings of the problem in this discussion as well.
Focus on the deeper goals driving the behaviours instead.
C2, for example, is a fundamental one that in its Napoleonic Warfare translation would prune entire branches (in the name of coherency, geometrical integrity, known course of actions, soldiers mood preservation, etc.).
It'd tell you to don't wheel at all if you have couple of friendly battalions falling shortly upon yours from the rear, becase that requires communication, coordination and reshaping the situation possibly under enemy fire. Hope you get the taste.
So think about how to prioritize these goals, look at utility theory, build compounds of elementary game tasks and then move your hierarchical searches forward in time, through planning (matching the abovementioned flanking manouver in advance, if possible).
GOAP may be a start point.
It's technically doable imho, if you don't care about savegames (you'd need to detour the save/load routine and handle your data serialization, in the case) and can put some effort into learning these techniques.
Mitra couldn't or didn't want to, unfortunately. He was tasked to deliver and Norb's pragmatism is known to be rather scaring. Maybe he just couldn't afford to move into even more vastly unknow fields under that pressure, who knows.
Best of luck.
r59

I've been modding the dll for quite awhile and the AI behavior is much improved. I agree the coding was poorly structured and even incomprehensible at times, but by rewriting it and making appropriate changes, the game gives a decent rendering of the napoleonic battlefield. Single player HITS even MP HITS is no so easy to win.