Those nutty tankers - strange AI behavior

[SgtMuhammed]

I have noticed a couple of times in almost every battle that some of my tanks seem to be doing doughnuts at the end of their movement orders. They go through multiple waypoints without trouble but go into orbit on the last one.

This happens in all types of terrain and most often without any enemy contact. It only happens with single vehicle movement orders.

Has anyone else seen anything like this?

[Bruno Weiss]

I had an opponent who's T-34 had a back and forth thing happening as it advanced towards some of my troops in a woods. It was kinda funny, well for me anyhow as it turned out on account of I had a 50mm AT gun trained on it as it was going back and forth. I don't speak German so I'm not sure what the lil gunners were saying, but I figure it was somefink like, Hold her still there Ivan till I get a bead on ya. Which, they eventually did. But yes I've seen some strange behavior on that note. But, I'd have to add the same sort of thing went on in CMBO even more often than it does with CMBB. So I figure we're better off now than we were.

[demoss]

I, too, have seen a T-34 who seemed to be auditioning for the Bolshoi.

Doug

[White Phosphorus]

One time an enemy Hertzer showed up and freaked the 4 34/85 into reversing without even targeting any of them. But then they immediately continued to move forward, then back, then forward. Maybe the Tac AI goes crazy when there is no cover to back up behind. (It was pretty much open steppe with a tiny tree patch back and to the right of my tanks).

[Laox]

This happened to me today! here is my example. I am Axis, that parboiled tank front and center was ordered to fast-move right next to its buddy..

This pic was snapped about 5 or 6 turns later in the scenario, after I had already created that russian scrapyard up on the road ..

Unfortunately, at the time of the parking request, I looked during the move, and to my horror the tank was doing some weird dance ahead of its designated spot, it was turning with stately grace anti-clockwise!! had the commader spotted a particularly rare flower perhaps? I know not..

Of course what happens? the first russian armor unit showed up coming in from the left, and took one look at the side of my tank and said I'll have that for breakfast, thank you. The damn thing even exploded, right as I was watching in disgust and thinking "noo! what are you DOING!"

Had it done what I told it to do, there would have been two guns trained on the road, with proper front armor facing the right way, and that russian scrapheap would not have cost me any tanks at all ...

As I say, there was nothing fancy here. The move for both tanks was Fast-move, Fast-move, Rotate to (face road). The last fast-move spots were next to each other, with ample space between them.

Could it be possible that there is a bug where the rotate sometimes decides to take the 340 degree trip rather than the 20 degree trip?

[Laox]

Oh one thought...

ergonomics on REVERSE vs ROTATE are not the best

The lines are the same color, almost, and the hotkey that comes to mind for both is "R". It is more common to "rotate" than "reverse" .. yet that is "O" (see what I mean about Ergonomics?), and "reverse" is "R".

So if you screw up R vs O, and do move-to, then "r" for rotate, you'll see a nice purple line, and everything looks good, but what you've really done is given a reverse order, which would cause a spin on the spot, before it was cancelled out next turn..

just a thought.

[Laox]

and - ah - speaking of ergonomics..

I HATE the way group select works to cancel out other orders.

Sometimes I am not aware I've got a bunch selected, then I concentrate on one unit and issue a command, and find all my previous work has been lost as all units selected get one new command, sometimes losing long path orders from the previous round(s) as well..

given the general hopelessness of group movement commands anyway, I'd rather that just be removed entirely and orders given can only be canceled by backspace, or halt etc.. individually, or, group-commands can only apply to units that have not currently got a command. Throwing away so much movement data without a confirm is something that should never happen!

sorry this is off this topic, but it relates to my last post above.

[Lt Bull]

Absolutely. I too have seen problems with tanks and waypoints and it is just pure lasiness why I haven't posted about it.

I have just run a few tests and have seen this problem again. It is easy to reproduce. Simply give a tank a move command straight ahead then make a 90 deg turn that extends for about 10m metres. I think what happens is that the game tries to "smooth out" the sharp corner of the 90 deg turn and the tank makes a wide turn. In doing so, the end waypoint ends up being on the inside of the turn so the tank ends up turning until it reaches the waypoint. What ends up happening is the tank ends up something more like 90deg to the direction you would expect it to be.

I gotta say that this is a major peeve for me and probably on the top of my list for it to be addressed. I have lost several tanks because of this behaviour. What compounds this problem is the need now in CMBB to try to minimise the number of waypoints to try to reduce command delay.

I would like to know whether this is a deliberate CMBB feature. I have checked with CMBO and the it seems that the vehicles in CMBO turm much more sharply than those in CMBB so they are less likely to end up at a wierd angle at the last waypoint.

The problem may be in other cases but definitely in the following case: If your last waypoint is 90deg to the waypoint before it and <15m in length, then your tank will end up facing 90deg from where you would expect it to be facing at the end of the move.

Dare I say, BTS fix or do somefink!

Lt Bull

[ January 07, 2003, 01:16 AM: Message edited by: Lt Bull ]

[Nidan1]

Originally posted by Bruno Weiss:

I had an opponent who's T-34 had a back and forth thing happening as it advanced towards some of my troops in a woods. It was kinda funny, well for me anyhow as it turned out on account of I had a 50mm AT gun trained on it as it was going back and forth. I don't speak German so I'm not sure what the lil gunners were saying, but I figure it was somefink like, Hold her still there Ivan till I get a bead on ya. Which, they eventually did. But yes I've seen some strange behavior on that note. But, I'd have to add the same sort of thing went on in CMBO even more often than it does with CMBB. So I figure we're better off now than we were.

Do you always kiss and tell, my friend.

[SuperSulo]

Originally posted by Lt Bull:

Absolutely. I too have seen problems with tanks and waypoints and it is just pure lasiness why I haven't posted about it.

I have just run a few tests and have seen this problem again. It is easy to reproduce. Simply give a tank a move command straight ahead then make a 90 deg turn that extends for about 10m metres. I think what happens is that the game tries to "smooth out" the sharp corner of the 90 deg turn and the tank makes a wide turn. In doing so, the end waypoint ends up being on the inside of the turn so the tank ends up turning until it reaches the waypoint. What ends up happening is the tank ends up something more like 90deg to the direction you would expect it to be.

I gotta say that this is a major peeve for me and probably on the top of my list for it to be addressed. I have lost several tanks because of this behaviour. What compounds this problem is the need now in CMBB to try to minimise the number of waypoints to try to reduce command delay.

I would like to know whether this is a deliberate CMBB feature. I have checked with CMBO and the it seems that the vehicles in CMBO turm much more sharply than those in CMBB so they are less likely to end up at a wierd angle at the last waypoint.

The problem may be in other cases but definitely in the following case: If your last waypoint is 90deg to the waypoint before it and <15m in length, then your tank will end up facing 90deg from where you would expect it to be facing at the end of the move.

Dare I say, BTS fix or do somefink!

Lt Bull

I had exactly the same happen to three tanks, twice, in my current battle. One single, long movement line, a 90° turn, and then a 10 meter move line "to get into position". They all ended up with the side to the enemy. This must be fixed (or explained, maybe we are doing something wrong?).

[demoss]

I have just run a few tests and have seen this problem again. It is easy to reproduce. Simply give a tank a move command straight ahead then make a 90 deg turn that extends for about 10m metres. I think what happens is that the game tries to "smooth out" the sharp corner of the 90 deg turn and the tank makes a wide turn.

I haven't seen this result in really odd facings, but I can confirm the smoothing out effect. In a recent game of [CENSORED], I had a number of vehicles widen their turns - right into some scattered trees. Fortunately, none bogged, but I was pretty aggravated by the slowdown.

Doug

[Lt Bull]

demoss,

The wide turns you are talking about is what I believe is causing the problems. Start a random QB, pick any side, there should be a vehicle there hopefully. Accept the defualt setup, give the vehicle movement orders: 20m directly ahead then another 10m left or right of the waypoint so the path is a 90deg turn. You will see what I mean.

The turn radius we see in CMBB for vehicles is definitely bigger than in CMBO which is resulting in this seemingly annoying vehicle movement behaviour. Any comments on this new feature BTS?

Lt Bull

[Guest Mike]

I just had a MG carrier charge around a wood into the fire of a MG bunker and 2 of those 3 x 15mm flak h/track thingies.....without orders from me of course!!

Does that count?

[Rother]

A 90 degree turn in 10m is a little tight for a tank moving at speed. In CMBO everythng did a pivot steer even vehicles that weren't capable. Looks like they finally fixed that. It should work just fine to do a rotate to order 90 degrees and then on the next turn move ahead 10m.

[DevilDog]

I've had this experience with my tanks several times. And it wasn't a mix up between rotate and reverse orders. It usually comes after a move command IIRC.

[Lt Bull]

Unfortunately Rother, putting a "Rotate to" marker at the end of the last waypoint doesnt really help and it is not a solution.

The pic below is an animated GIF that shows six instances of T-34s been given a "fast" move waypoint directly ahead, then another waypoint approxiamately 10m-13m to the right at 90deg.

(image replaced by link to lower thread loading times)

The placement of the waypoints seem innocent, resonable enough and easy to understand. However CMBB's "path finding" interpretation of them is nothing like what anyone would reasonably expect to occur.

So what's the deal? Is this intentional? If so is there a sure fire way to avoid this obviously undesirable situation. I have noticed that smaller vehicles have tighter turning radii than the larger vehicles. Perhaps a list of turn radii is needed so you can check each vehicle, though this would indeed be tedious.

As an aside for future game design reference, perhaps when plotting waypoints in the orders turn for vehicles, the vehicles minimum turn radius is lightly superimposed when plotting waypoints. It would be similar to how some CAD programs fillet adjoining lines at the vertices. You could see whether a vehicle could make a turn or not.

Lt Bull

[ January 08, 2003, 10:34 AM: Message edited by: Lt Bull ]

[SgtMuhammed]

Lt Bull

I think you are dead on about the turn radius. Maybe the path indicator could change color if a point is within the turn radius much like the LOS or Targeting line. While not allowing everything to turn on a dime is more realistic there has to be a better way of modeling the turn radius than having vehicles endanger themselves needlessly. After all even if we don't know the turn radius of a KV one would assume that at least the crew does.

[Alsatian]

Lt. Bull,

How about running the same test with one modification.

Instead of ordering the tanks directly forward, order them at an angle so they all shift over one "column" to the right. That would make the 90 degree turn at the end of the movement a little softer.

Do they still do the tanker two-step?

Guess I'll have to dust off my manual from driver's ed and touch up on my K-turn skills.

[Lt Bull]

Alsatian, in thinking about what you suggested I came up with another setup.

To demonstrate the fine line between what you would call a "successful" execution of tank movement orders and what is (has been) potentially disasterous execution of tank movement orders movement:

On the far left is approximately a 3m right 90deg turn. The next is approximately 7m, then 10m, 13m, 17m and 20m.

Note how moving the waypoint just a few metres more (10m to 13m) can mean the difference between potential disaster and what you would expect.

Sgtgoody, I like your idea of having the path indicator change colour if the waypoint is placed/positioned inside that tanks/vehicles turning circle. A great idea and seemingly simple.

I might further investigate movement at angles less than 90deg (as Alsatian suggested) and see what the limitations are for those.

To lower loading times for this thread, I have made the original GIF file a link. View it here.

Lt Bull

[ January 08, 2003, 11:24 AM: Message edited by: Lt Bull ]

[MikeyD]

A 90 degree turn while moving fast isn't a particularly good idea, apparently. Looks like the tanks are trying to get back to the waypoint they overshot. Have you tried the same thing at a 'move' or 'hunt' speed?

Sure this may be a bug, but I'm actually kind'a impressed. I didn't know CMBB had incorporate turn radius info into the vehicle movement model.

[Lt Bull]

Doing a 90deg turn with a tank in CMBO at FAST speed presented no problems. I have conducted the same test with Shermans in CMBO and the results can be seen below. In both CMBO and CMBB the tanks actually slow down when taking the turn. The way the corner is negotiated by the Sherman does not seem to be "unrealistic" and it does not seem to be "turning on a dime" in this movement.

It is obvious that "something" has changed since CMBO about the way vehicle movement paths are calculated. Unfortunatley many of us have found this out the hard way....in the heat of a battle The consequences of being caught out by this new apparently undocumented change can be disasterous and simply baffling to the unwary.

I understand the concept of a "turn radius" may be considered more "realistic" but it is definitely not realistic to see your tanks do that spin at the end of their move.

I would really like to know what the idea was behind this change in waypoint/path tracking for vehicles in CMBB. The way it was in CMBO did not seem like it was begging to be changed.

BTW, I have conducted idetical tests with the T34 using MOVE and HUNT commands and I did not see the "spinning" behaviour you see when using the FAST command. This perhaps indicates that the "turn radius" for tanks is smaller when moving at speeds other than FAST. Although technically "realistic", all tanks in CM slow down when they are taking a corner and I would go as far to say that the corner speeds for FAST and MOVE are not that much different, hence making the apparently larger "turn radius" imposed when taking a 90deg corner somewhat inappropriate.

Would really like to clear this one up.

Lt Bull

[Ex Bellator]

I'd like to congratulate Lt Bull on his work with this problem. I hope you get some deserved feedback from BFC soon.

[Joques]

AH! I just had this happen to me. A Stug, orders were fast move ahead followed by a 90 degree turn right. It turned the 90 degrees, and then 90 degrees more! Very unnerving, esp. since there is a Russian heavy lurking just beyond LOS, and my Stug just turned its flank towards it!

In fact, I'll send the relevant PBEM files to Battlefront. Which mailadress should I use?

[Lt Bull]

Originally posted by Joques:

AH! I just had this happen to me. A Stug, orders were fast move ahead followed by a 90 degree turn right.

Consider yourself one of the lucky ones. Your tank and tankers survived. But there have been many before you and many more still to come for whom it will end in tears, and in a burning, twisted, smoking heap of scrap metal that doubles up as a coffin, being left only with one question..."Why? For what? Those damn nutty tankers!"

BTS, please address this sensless slaughter before it takes more innocent victims.

Lt Bull

[Lt Bull]

(burmp)

be sure to let the animated gifs in the posts above load