Problems in CMBN V3.11

[Vanir Ausf B]

Judging from the screenshots I believe this is a bug that was introduced in 3.00. Or at least I think it's a bug since the pathfinding AI is much different in v2.12 in the same situation. As someone else said, in 3.00 you need to place a waypoint very close to the other side of the gap in order for the AI to use it. In 2.12 the AI would recognize the gap as the shortest route regardless of where the waypoint is placed (I have a copy of v2.12 on my comp so I can compare).  I reported this almost a year ago so I don't know when or if it will get fixed.

 

It's hard to judge what is going on with the Wespe without seeing the hit text. For example, you noted one hit on the "gun assembly" but without seeing the hit text we don't know what part of the vehicle the game engine considers that spot on the vehicle to be part of.

 

I don't see anything to suggest there is definitely an issue with penetration lethality, but I may take a look at it anyways. There is an issue related to that with T-34s in CMRT that is getting fixed in the next patch. The problem is that there isn't much real world data on what to expect so it mostly comes down to gut feeling.

 

Hey guys!

 

Okay, I guess I was wrong with the pathing behaviour.

I think about some testing though.

 

But to give you an impression how I came up with the idea, here are some more screenshots.

 

Green line - intended way, set with two 'quick' waypoints, one half way across the field, roughly on the green arrow the other one at the end of the field.

Red line - the way they took, there was an hMG 42 down the road, which I knew is there but not the AI (no contact icon)

 

What I don't understand is why the AI didn't choose the passage in the Bocage, which obviously was the shorter and easier way (no gate to jump over).

 

Result. The guys at the passageway are about to start their dash (20 seconds delay).

 

Waypiont of the second team. Looks like the AI sticks to roads tenaciously.

 

This is not to press my piont, but just to show you how I came to the conclusion there is something odd with the pathing.

Hope this shows you that my intention was not to troll.

Edited July 29, 2015 by Vanir Ausf B

[Battlefront.com]

Judging from the screenshots I believe this is a bug that was introduced in 3.00. Or at least I think it's a bug since the pathfinding AI is much different in v2.12 in the same situation. As someone else said, in 3.00 you need to place a waypoint very close to the other side of the gap in order for the AI to use it. In 2.12 the AI would recognize the gap as the shortest route regardless of where the waypoint is placed (I have a copy of v2.12 on my comp so I can compare).  I reported this almost a year ago so I don't know when or if it will get fixed.

My memory is fuzzy on this one, but it could be that this was intentional. The small gaps are difficult to see unless you get in close. I seem to recall units were winding up breaking into the field earlier than expected because of gaps they didn't notice.

Whatever the case is, this is one of those situations that it has to be one way or the other and neither is going to be correct all the time.

Personally, I think it's better to have a system which offers the ability to "instruct" the AI as to your explicit intentions. The only way to do that is through waypoint placement. The way it works now allows that, the way it worked before did not. So on the surface I'd say the current way is superior to the old way. Though, of course, there is plenty of room for disagreement.

Steve

[Vanir Ausf B]

My memory is fuzzy on this one, but it could be that this was intentional. The small gaps are difficult to see unless you get in close. I seem to recall units were winding up breaking into the field earlier than expected because of gaps they didn't notice

 

Yes, that's a good point. I've had that happen to me a time or two. But if people don't know about it frustration can result and there is no indication that small bocage gaps are an exception to the general rule that units take the shortest available path.

 

Aside from that my personal objection to it is that it requires an extra waypoint between bocage lines. Infantry pause at each waypoint and that is not a good place to be pausing if there are enemy units behind the next bocage line.

Edited July 30, 2015 by Vanir Ausf B

[RockinHarry]

Think that infantry taking always the quickest route, is not quite true. From my testing if there´s certain cover AS between the waypoints, then the Tac AI will oftenly take a longer route by using the cover underways. In DasMorbos´s screenshot the road offers more "cover", than going straight over the field, by using the gap through the bocage!

As I already mentioned previously, the AI has a very high liking for "road ditches", as well as craters, no matter how small they are (or look on the map). They´re almost like magnets to the AI and this occasionally leads to the same problem with foxholes, when the pixeltroopers don´t all move in and take positions outside the individual foxholes. Have a look at (small, almost invisible) craters around the foxholes, even if in ajacent AS and I bet some the ptroopers select these as better "cover".

This can either be used to advantage (if considering this behavior), or avoided, particularly when setting up an AI opponent.

If BFC is motivated to look into something, then maybe the "weighting" between certain cover possibilities. Road ditches, craters, but also ditch locked AS (either + or -) would be the candidates.

With regard to desired movement route and tactics in the screenshot example (infantry through bocage gap) and when in doubt about possible route chosen by the AI, then it would´ve been better to stop at the gap, split off some scouts and send them off first to see which route would be selected, particularly if some enemy presence toward next movement point is known or suspected. That´s all in the hand of the human player to deal with, but extremely difficult to plan with AI players actions.

Also the bunching up happens more oftenly, when the start movement AS already has the ptroopers bunched up, or generally when individual squads are too close to each other. Dependent upon squad size and number of team composition, start and end movement AS need to be considered for each individual team as well. Assigning a face order on the last movement point, gives an indication where a squads teams will end up. If the route is not too long, individual teams will move parallel to the next movement point. If it´s longer, the teams will form a squad column at a certain range and fan out again when reaching the final AS. Otherwise, using too many movement points at too short ranges has the disadvantage, that the squads need time to deploy at each individual movement point, increasing exposure and such. Yet need to find the "green range" for movement lengths, avoiding the most serious troubles, when not moving a squads teams individually.

Edited July 30, 2015 by RockinHarry

[Vanir Ausf B]

I have tested this on a different map (Sheriff of Oosterbeek). It is not cover magnetism at work. Example:



But it's not consistent. In some places units will go through the small gaps on their own. One factor seems to be if one or more of the ends of the bocage line is open. If it is they will run around that open end and ignore the gap unless you place a waypoint with 2 action spots of the other side of the gap.
 
But if both ends of the line of bocage are anchored on an impassible terrain feature such as another line of bocage the units will not run backwards to find a way around. They will use the small gap even if the next waypoint is far away.
 
For example:


But this isn't 100% consistent either. In the example below the unit will use the small gap instead of the open end even with the long movement order.

 

[akd]

On the Panthers, were you perhaps playing with high morale setting on the tank crews? From your screenshots, crew behavior seems to be problem, if there is one. You've got crew sitting in tanks that in reality would have been abandoned and considered KO'd.

[DasMorbo]

About the Panthers -  All were buttoned up when hit.

Shots were fired at less than 400m/yards, mostly on side at sides with thin armor.

 

 

A little physical examination of whats going on when a tank is hit.

Yes, the different warheads have differing physical values, but I am using the Fermi-principle here, showing in which magnitudes the physical processes take place.

 

So lets take a hit from a PIAT: it penetrates with an jet of molten metal, travelling at about 7-10 km/s and a pressure of no less than 100 bar (up to 200 bar or 200 GPa) and a temperature of at least 350°C and up to 480°C.

Unless the tank has it's hatches open, so the immense overpressure can dissipate, the crew will take seriuos injuries to all presure sensitive organs - lungs rupture, hearing and balanceing organs will take serious damage and so forth.

If that is not enough, the super fast jet creates a cloud of molten and shattered steel which tends to create severe injuries and fires upon impact. And don't forget the very hot gases.

 

Look at AP shot penetration: if a solid 57mm (6pd) AP shot penetrates the armour, it still has considerable kinetic energy. I have no numbers aviable, but considering the AT guns of the time had muzzle velocities of 600-1000 m/s (meters are roughly yards). Lets consider the amount of energy lost through flight and penetration, I would expect an AP slug fired at less than 300m to bounce around inside the tank at 100-200 m/s. The 75mm round had an explosive filler, so it created even more damage if it went off.

 

 

Looking at the punishment the Panthers took, I seriuosly doubt everything is okay here. My first impression was that for some reason ammo and fuel are not correctly simulated. I derived this conclusion from the fact that out of seven PzVs only one burned and only after recieving fice solid penetrations. No ammo explosions at all. But I could be wrong.

 

Maybe it is not a bug, but then I would have to say this is unrealistic.

Not even an veteran tank crew would be able to keep fighting after receiving more than two hollow charge penetrations into an closed fighting compartment, most likely the first would have wounded or killed everyone.

Same goes for the amount of AP penetrations received, it showed pecuilar few effects, considering the physical forces at work and the amount of flammable and explosive materials inside a Panther. Just look at the inside model in-game.

[Battlefront.com]

I have tested this on a different map (Sheriff of Oosterbeek). It is not cover magnetism at work.

Yup. The TacAI has a lot of flexibility depending on a wide variety of variables. It is easy to generalize behaviors, but because there are so many diverse factors at work it's absolutely impossible to have it all boil down neatly to a simple explanation.

For example, in your last screenshot with the long movement, this makes sense when you take into account that the AI is aware of how close the unit is to the gap at the time the order is started. In this case it appears the AI "guessed" your intention was to move through this gap because it is so close by. But if you had your unit, say, 50m distant it might not have. At least that's how I remember things and your various examples above seem to confirm it.

I definitely remember a LOT of complaints about gaps resulting in infantry going into fields in unexpected places/times. I am sure that the changes that took place were specifically designed to minimize that sort of thing. As imperfect as it may be

Steve

[Battlefront.com]

A little physical examination of whats going on when a tank is hit.

Yes, the different warheads have differing physical values, but I am using the Fermi-principle here, showing in which magnitudes the physical processes take place.

 

So lets take a hit from a PIAT: it penetrates with an jet of molten metal, travelling at about 7-10 km/s and a pressure of no less than 100 bar (up to 200 bar or 200 GPa) and a temperature of at least 350°C and up to 480°C.

Unless the tank has it's hatches open, so the immense overpressure can dissipate, the crew will take seriuos injuries to all presure sensitive organs - lungs rupture, hearing and balanceing organs will take serious damage and so forth.

If that is not enough, the super fast jet creates a cloud of molten and shattered steel which tends to create severe injuries and fires upon impact. And don't forget the very hot gases.

Where the hits are registered is important, as is the degree of reliance you have upon theory vs. reality. I say the latter because it's been shown, time and time again, that purpose made thermobaric weapons on modern battlefields often do not produce the effects that one would predict from such a weapon. The reason being that overpressure is highly sensitive to specific conditions.

I'm not saying that everything in CM is perfect, rather that the theories applied to it sometimes are themselves imperfect.

 

Look at AP shot penetration: if a solid 57mm (6pd) AP shot penetrates the armour, it still has considerable kinetic energy. I have no numbers aviable, but considering the AT guns of the time had muzzle velocities of 600-1000 m/s (meters are roughly yards). Lets consider the amount of energy lost through flight and penetration, I would expect an AP slug fired at less than 300m to bounce around inside the tank at 100-200 m/s. The 75mm round had an explosive filler, so it created even more damage if it went off.

All of this is explicitly simulated. I can promise you that if there is a significant problem with the way the game is coded there would be ample evidence of this and we would have had our chestnuts roasted on an open fire years ago. This and the Tiger are the two most commonly nitpicked vehicles in the game.

 

Looking at the punishment the Panthers took, I seriuosly doubt everything is okay here. My first impression was that for some reason ammo and fuel are not correctly simulated. I derived this conclusion from the fact that out of seven PzVs only one burned and only after recieving fice solid penetrations. No ammo explosions at all. But I could be wrong.

Glad to hear that you are considering that possibility There are several dozen people, like you, trying to find major flaws with the game at any given time. The fact that they haven't found anything significant for quite a long time is certainly not a result of a lack of effort. At this point I do feel quite confident that if someone thinks they have found a "fundamental flaw", that it is unlikely to be true. I say this because we have had people beating this thing to death for 7 years now. You guys are not that slow to find problem areas. In fact, I'd say you are depressingly efficient at finding flaws!

 

Steve

[A Canadian Cat]

I have tested this on a different map (Sheriff of Oosterbeek). It is not cover magnetism at work. Example:

Very interesting and enlightening @Vanir Ausf B look at what happens and that there was in fact a change. Good to know.

 

I learned early that to get what I wanted I needed to help out. I agree with others that putting a way point on the far side of the gap in the field is not good from the point of few of pausing in the open.  So instead what I do is put a way point on the safe side of the cover right in front of the place I want the men to go through:

 

 

 

I have been doing that for a while now and it has been working for me.  Now looking at that last example where the team goes through the gate in 3.x  I kinda wonder if my technique would break.  I have been consistently doing this since the early days and I have not been surprised but this exact scenario you have above makes me wonder.  I'm going to have to try it to make sure.

 

[General Jack Ripper]

 

 

 

 

That's exactly how I place my waypoints as well, and 99% of the time it works as intended. The only time it hasn't worked is if the second waypoint for the destination was offset to some degree. For example, if the destination waypoint was 45 degrees to the right of the one pictured, your pixeltruppen might decide to run down to the open end of the hedgerow anyway.

You have to think of movement paths the way it would work in real life. You wouldn't simply tell your troops "move over there", you would give them a specific instruction, "get into that ditch and follow it to the other side of the road", etc. You simulate specific instructions by drawing a path of waypoints to your destination.

[c3k]

What IanL said.

 

VaB: very cool "time lapse" image.

[womble]

While we're on pathing, I've had a problem lately with engineers blowing holes in hedges. Sometimes, if there's a gap nearby, they'll run round through the gap rather than Blasting through the hedge they're adjacent to. Worked fine in some places, not so well in another. The exception was a diagonal rather than orthagonal hedge. I can probably dig the save out (it's in KG Engel, mission 1) if it's of interest.

[Vanir Ausf B]

I have been doing that for a while now and it has been working for me.  Now looking at that last example where the team goes through the gate in 3.x  I kinda wonder if my technique would break.

It would

 

 

 

The team still goes to the end of the bocage line, the difference being that two soldiers go through the gap first so the team is effectively split until they meet up at the fence.

 

But this doesn't happen in all instances. Example 2:

 

 

Here the team completely ignores the gap, same as if there was no waypoint next to it.

[A Canadian Cat]

I have tested this on a different map (Sheriff of Oosterbeek). It is not cover magnetism at work. Example:

I just tried the same thing. I got the same results giving the orders like Vanir and my way. Two guys went through the gap and two guys went around the end through the gate.

 

I ran the test twice with each order - same results all four times.

 

I ran one more where I gave a 5s pause and a face command at the gap. That time one guy went through and three went around.  ???

[A Canadian Cat]

LOL cross posted similar results.  I don't think I like this.  Good thing I have not run into it much.

[Vanir Ausf B]

The more I mess around with this the more inconsistent and unpredictable it gets.
 
Going back to our favorite Oosterbeek gap, if I plot a waypoint directly on the other side of the gap the team uses the gap. This is the expected 3.0+ behavior.
 

 


Now that our team is at its destination I plot the same movement in reverse.
 

 

 

As you can see, in this situation the waypoint on the opposite side of the gap doesn't prevent them from taking the scenic route.

[sburke]

The more I mess around with this the more inconsistent and unpredictable it gets.

 

Going back to our favorite Oosterbeek gap, if I plot a waypoint directly on the other side of the gap the team uses the gap. This is the expected 3.0+ behavior.

 

 

Now that our team is at its destination I plot the same movement in reverse.

 

 

 

As you can see, in this situation the waypoint on the opposite side of the gap doesn't prevent them from taking the scenic route.

That is the reverse in route, but perhaps not the opposite in actual steps.  Plot the return by making the first way point directly on the units side of the hedgerow, not the opposite side.  The second waypoint to the final destination.  What you did looks the opposite, but I think may actually a different plot from the TAC AI's viewpoint.  I suspect if you plot the waypoints exactly the way you did in the first move coming back then they would use the hedgerow gap.  The distance from the start point to the location opposite the gap is probably faster from the AI's view by going around.  You didn't quite give it that option the first time because it's first waypoint is directly on the opposite side of the hedgerow.

Edited July 30, 2015 by sburke

[RockinHarry]

Btw, what, if any, influence has the axis/allied friendly map edge setting on pathing near or through bocage, walls and the like? I seem to remember something...

Beside that, I agree with Sburke concerning Vanir Ausf B´s plotted movements forth and back, which are not the same. Also things get even more intersting, when moving whole squads at or through bocage. It matters i.e if either the squad leader team or the support teams occupy the AS directly at the gap. Quick move, usually gets the squad leader team move first, while Assault move more oftenly than desired, gets the support team move first by the Tac AI.

[Vanir Ausf B]

That is the reverse in route, but perhaps not the opposite in actual steps.  Plot the return by making the first way point directly on the units side of the hedgerow, not the opposite side.  The second waypoint to the final destination.  What you did looks the opposite, but I think may actually a different plot from the TAC AI's viewpoint.  I suspect if you plot the waypoints exactly the way you did in the first move coming back then they would use the hedgerow gap.

 

Yes, I was trying to keep to the rule that you have to plot a waypoint near the opposite side of the gap either way. You are correct that if I plot the exact same waypoints the team uses the gap in both directions. I suspect this is because the end waypoint on the way back is close enough to the gap to act as a "take the gap" que.

 

But that changes if you lengthen the starting leg so that the destination waypoints are closer to the same distance on both sides of the bocage.

 

 

 

So far so good, as long as you're ok with pausing on the exposed side of cover

 

Return trip with identical waypoint placement:

 

 

Not ideal.

[Battlefront.com]

You have to think of movement paths the way it would work in real life. You wouldn't simply tell your troops "move over there", you would give them a specific instruction, "get into that ditch and follow it to the other side of the road", etc. You simulate specific instructions by drawing a path of waypoints to your destination.

 

Yup. There is rarely a single "right way" to do something. The less specific your instruction, the wider the variety of choices are available. The wider the array of choices, the more diverse/unexpected the results. It doesn't matter if you're talking about the pathing AI or issuing instructions to an underling in a crime syndicate A lack of specificity is likely to result in FUBAR.

 

I just tried the same thing. I got the same results giving the orders like Vanir and my way. Two guys went through the gap and two guys went around the end through the gate.

 

You know, I think I remember this specific map and that specific part of the map as being problematic. From what I remember, it has a bunch of pretty twitchy conditions in a very small space. Not a good combo. Like telling someone to "cut the red wire" to defuse a bomb like this:

Not as likely to wind up with a good result as a bomb like this:

 

What you did looks the opposite, but I think may actually a different plot from the TAC AI's viewpoint.

Absolutely. Some weights are cumulative and are affected by other factors, such as distance. What looks like a straight forward path from one direction does not necessarily look the same in reverse. Add to this that you might not be clicking on exactly the same place within the same Action Spot each time. If you click 1m to the left maybe that's just enough to start the path to the left, but 1m to the right makes it start to the right. Since a path is a series of decisions, one small difference can result in an entirely different path. Which is why we should hope nobody ever invents a time machine!

You guys can think about it this way. You're in an office building and there's two pretty similar paths to get to the kitchen. You are instructed to go there, by your boss, to get a doughnut and you now he wants it fast. At the moment you are given the instruction you happen to be facing Path A, so you take it because you don't have to change direction and it's pretty similar to Path B in terms of speed, obstacles, and the hot women you like to oogle over (and yes, they all notice and think you're a creep. But that's a different lesson . After you get the donut you turn around and find yourself facing Path B. You think about going back through Path A, but then you think that you didn't see any of the women you're thinking of asking out (aka stalking) along Path A, so you decide to instead go down Path B. Either way you get to where you need to be in a timely manner, so why not go Path B this time?

This is how the AI thinks, though without the creepy leering and awkward comments about how nice a woman's feet look. Unless the Boss explicitly states "go down Path A", perhaps because Human Resources has already warned you that Beth feels uncomfortable about you, then you get to decide which way to go based on your own internal decision making methodology.

Or something like that.

Steve

[Battlefront.com]

Not ideal.

Depends. Notice that by lengthening the line you add a new variable... trees! Trees have an impact on pathing decisions because they affect movement and cover. By lengthening the line you are now including the trees in its decision making process. I am guessing that cutting through the trees was just enough of a factor to influence the AI to try and path around them. In that case, going around the trees means not going through the gap. Which then means the path is "ideal" using that logic.

The problem is there's millions of possible combinations of factors and there's often more "wrong" ways in the player's mind than there is in the AI's. Since it can't know what your intentions are beyond waypoints and Order type, it has to make an educated guess based on GENERALIZED logic. There's really no way to avoid these sorts of situations completely, though of course tweaking the equations sometimes does help out in some situations. Unfortunately, it also can lead to other situations being worsened. Such is the double edged sword of changing instructions. Real life has the same thing. You see someone doing something stupid and you say "NEVER EVER do that!". Then a situation where the action actually makes sense, or is the best of the worst choices, the instruction suddenly becomes a problem even though 98% of the time it's really sound advice.

Steve

[Vanir Ausf B]

Getting back to my previous point, even when you place a waypoint directly on the other side of the gap that doesn't guarantee the gap will be used.
 

 

 
???
 
It seems that the only way to ensure a gap will be used is to place waypoints near the gap on both sides.
 

Edited July 31, 2015 by Vanir Ausf B

[Battlefront.com]

Yup, it's pretty obvious you found a situation that is particularly tricky for the AI. As I said, I remember this very specific map section from earlier testing days. It's something that stresses the system, therefore the range of results reflects that state.

I can't say this enough... pathing is not an exact, simplistic cause/effect. There's far too many variables to take into account for it to be anything but simplistic. The more complex something becomes, the more varied the outcomes will be. Those looking for simplicity will be disappointed since it doesn't exist. And that means it is absolutely impossible to say, with any degree of absolute assurances, that a SPECIFIC course of action will work for ALL situations ALL the time.

Now, is there something that Charles might be able to tweak here? Perhaps, perhaps not. It is also possible that making a tweak to achieve more predicable ("desired") results in this situation will produce less predictable ("undesirable") outcomes elsewhere since everything is by necessity interconnected. There is no such thing as an isolated branch of the AI pathing that is specific to gaps. Gaps are simply a variable in a mix of dozens and dozens of other variables.

As I've said, pathing will always be twitchy and imperfect to some degree. You have shown us an example of a case where that is true. You can no doubt find others by focusing on gaps or something else that is a chokepoint (the most difficult of all terrain features for the pathing AI to deal with). However, without also documenting all the "correct" things the pathing AI does then we won't have context. Having 1 in 10 pathing decisions go wrong is very different to us than 1 in 100 or 1 in 1000. Based on personal experience with the game and with those playing it, I'm confident that the ratio of good to bad decisions is lopsided in favor of the good. If it wasn't, we'd see thread after thread complaining about pathing and there being no real good advice to mitigate it. I know this for sure because there definitely were times when this sort of discussion happened. As I've already said many times, you guys aren't shy about telling us when you think something is broken

Steve

[RockinHarry]

Getting back to my previous point, even when you place a waypoint directly on the other side of the gap that doesn't guarantee the gap will be used.

 

 

 

???

 

It seems that the only way to ensure a gap will be used is to place waypoints near the gap on both sides.

 

Here´s the view of the particular map part in the editor. The screenies from previous examples (Sheriff of Oosterbeek) have view from north to south (up the screen). 2D Editor is south to north. Allied friendly map edge setting is west, axis is east. Map edge settings do not have any influence in this example. Ok. Cover terrain between plotted movement neither. Almost always the team will take the route across the wooden fence until...waypoints are plotted 1 AS radius of the gap in the bocage OR waypoints directly in the AS with the gap, which in fact requires 2 waypoints within the same AS . In this case the AI will most realiably take the route through the gap. Tested with a 4 man, split off team and Quick move only.

It also doesn´t matter to remove the brush from the bocage gap, or remove the road between the movement points.

I do not draw conclusions. It´s just observation and how to tackle movements through the terrain types bocage, high walls... with gaps. I´ll also take this into consideration when setting up zones for the AIP, when there´s any such terrain types on map.