Pathfinding Issues?

[Big Poppa Pump]

Playing Street Sweep, WEGO, in 1.08. I order a two man team to fall back behind a wall and into a building a few meters behind them. Hit end turn and watch the results.

The two man team splits with one man running clockwise and the other counterclockwise around a burning car in the opposite direction of the ultimate destination. They meet in the middle of no mans land, pause, then one man retreats back to the building (as per the original order)while the other circles the car one more time before completing the order. Of note one man appeared to enter and exit a nearby building without using doors (maybe a clipping issue). A save game is available.

I'm guessing this has something to do with the complexity of the map or perhas something to do with nearby vehicles. Earlier I ordered a fireteam to relocate to the other side of a wall with a friendly tank nearby. The fireteam did something similar around the tank. One man looped around the tank (into the middle of the street) and the other seemed to slide around the wall through a nearby building (not through a door).

Like I said I think this may be due to complexity of the map design as I have found that in most instances pathfinding is very good but these two events stood out as noteworthy.

[JohnO]

BigPoppaPump, if you want to send the file to me you can. I will see what happens and post it in the Skunkworks. I will post what we/I find as soon as possible, ok. Thats all I can do just bring the issue up the ladder. Also make sure you have CM:SF in the subject, ok.

[Big Poppa Pump]

File sent.

[JohnO]

Got it. Will look at it first thing in the morning, getting late here and I'm tired.

[Big Poppa Pump]

Thanks. I was also able to recreate the other instance. Sent two additional files to you.

[JohnO]

Got them, sent a reply back to you about the weired things that I saw as well. Will get back to you as soon as possibly.

[JohnO]

Just to keep you posted, the files are uploaded and waiting for a reply.

[Dirtweasle]

Played a couple CMSF scenarios in 1.08 last night and there is still something off about the path-finding. Main problem is it's unpredictable. Where the squad will actually go and if it will actual stop on the point you place seems to make little sense. For instance; I had a squad get killed because it decided to go the long way around a group of buildings instead of through. Another, they were supposed to run and jump over a little wall and take cover on the other side which did not happen.

[Thomm]

Originally posted by Dirtweasle:

Played a couple CMSF scenarios in 1.08 last night and there is still something off about the path-finding.

Fully agree with your points!

Discussions about potential improvements are ongoing!

Please do not hesitate to send me save games that show unusual behavior! I shall make sure that they are stored appropriately.

Best regards,

Thommm

[Dirtweasle]

I didn't save any games, but will share these observations which I x-posted at another gaming website regarding unpredictable endpoints and path-finding issues -

Other than in a building, which the squads predictably occupy at the end of a waypoint time after time. In the open, meaning not in a building of any sort, it's very unpredictable how a squad will get from point A to point B, and/or precisely where the squad or other unit will actually end up.

Sneaking squads around, rushing past open spots, finding cover and concealment, leapfrogging squads is practically impossible for me in CMSF, and this is where CMx1 while it was while a little off from reality what with the squares and all actually was a much better game. A much more enjoyable time. I played 1.08 for a several hours this weekend and last night for maybe 45 minutes before giving up on it again so it's not for a lack of trying.

I hope the WWII offering works out.

In the mean time you guys really need to test the heck out of all things unit pathing. I was again seeing squads go off at a 20 degree angle to start a path when the intial waypoint was right dead nuts straight ahead.

When a squad ends it's plotted movement orders, other than in a house, it's quite literally a crap shoot where it will actually end up within like a 20 - 30 meter area. That's not good enough guys. Not good enough at all for a game with a 1:1 high fidelity system. Sorry, but there you have it.

[gibsonm]

Well then I guess 1.09 will be out in say 2056?

[Dirtweasle]

If you say so I suppose so.

Right now maneuvering units it's a lot like driving an old pick-up truck with a sticky accelerator, bad brakes, a loose steering wheel and bald tires.

Look, I don't know what more to say, the pathing is messed up. I'm very disappointed that it is, I wish it were not. If it's un-fixable any further in a reasonable length of time well that is good to know too.

[thewood]

Sarcasm seems to the in style response now.

[gibsonm]

Well as I see it you have three choices:

1. “In the mean time you guys really need to test the heck out of all things unit pathing“

That means the delays between 1.08 and 1.09 blow out exponentially because we testers do it in our spare time (you know when we aren’t doing our normal jobs, etc.). Part of the reason why I’m typing this at 0100 local time.

Then a mass of people complain as to why its taking sooo long for the patch to come out.

2. We allocate a “reasonable” amount of time until Steve / Charles are happy and release patches more frequently and accept the fact that it will be an evolutionary process.

3. You pay a hell of a lot more for the product so Steve / Charles, et al can pay some people to test full time and you get the game earlier.

Your choice I guess.

If it sounds sarcastic, sorry but it seems amazing to me that people haven’t latched onto this yet.

[JohnO]

I posted the save game files to the projects for Charles to look at last week and hasn't response to the questions I asked him. So I took the liberty of sending them straight to Steve to send to Charles to look at the pathfindings problems that BigPoppaPump has found. Steve sent a reply back that Charles was busy coding the Marine Module and stated that he forward it to Charles for me. I should have something back by this weekend I hope to post here for Bigpoppapump. So the issues about pathfinding is being looked at. I must admit I'm not a big fan of MOUT fighting, I like open terrain for my battles.

[Thomm]

Originally posted by Dirtweasle:

When a squad ends it's plotted movement orders, other than in a house, it's quite literally a crap shoot where it will actually end up within like a 20 - 30 meter area. That's not good enough guys. Not good enough at all for a game with a 1:1 high fidelity system. Sorry, but there you have it.

Again, I agree with you, and I shall try to get the point across!

I strongly hope that the remaining pathfinding issues will be fixed! I am pretty confident that they will be. When I look at the game and how many facets of it are working already, some necessary improvements appear to be become feasible.

Best regards,

Thomm

[Dirtweasle]

Thanks Rollstoy. A lot has been done already. The frustrations from the UI, hotkeys, etc... and all the other nits are about picked over. There's a lot going on in the game too, a lot to look at during a turn. Right now in 1.08 the obvious sore thumb sticking up that I see is pathing.

Has there been documentation provided surrounding why units so often miss the waypoint, or stop short, or veer off to the side of one? The thing that gets me is I cannot figure out why it happens now.

For instance; I get it now about vehicles moving at full speed down a road. You need to slow them up. Like in real life tm you got to prepare for a turn. I got it, and added it to my playing style, my SOP when when plotting a path down a road now includes managing the placement of the extra "slow" waypoints. But.... I cannot figure how to manage what goes on with the pathing; it seems to me just random. The overshooting and undershooting of the waypoints defies any sense I can appreciate by looking at the game screen. hence it's frustrating.

Thanks ayways Gibson, but I'm not looking for an argument on what to accept "for the good of the cause" after all this time.

[Thomm]

Originally posted by Dirtweasle:

Has there been documentation provided surrounding why units so often miss the waypoint, or stop short, or veer off to the side of one? The thing that gets me is I cannot figure out why it happens now.

Cannot answer, NDA is kicking in!

I take the liberty, though, of writing that some days ago I submitted a test scenario that highlights precisely the issue that you write about.

Best regards,

Thomm

[Dirtweasle]

Well, if any of the official BFC crew are reading this an explanation you could share with the customers that would make the game more satisfying would be very nice at this point. Including your plans or intentions on dealing with the matter would be another good step.

[Battlefront.com]

Dirtweasle,

We comment about stuff like this all the time, so not sure why you think we've suddenly lost interest in this topic. Go back and look at the release notes of all patches and I think it should be very clearly evident that we're interested in eliminating as many pathfinding issues as we can identify. Several testers have also posted, who in turn post to internal discussions (which you testers are getting a wee bit too close to bringing out here... so knock it back a bit please). You've even been told that save games have been sent off to Charles and myself to check out. So I guess I'm not sure why you feel this issue is just floating around without any concern about it.

The fact is that pathfinding is a problem for just about every game ever made. Why we should be expected to get 100% of all paths 100% correct 100% of the time is... rather unreasonable. I've played games with binary terrain that the game system can't handle, and yet that game sold something like a million units at a time when the average "hit" sold 100,000. This is not an excuse for pathfinding issues in CM:SF, it is simply a bit of perspective that would be helpful if you had vs. a theoretical expectation that is likely to never be met.

With that said...

Sometimes pathfinding gets blamed for doing things that it should do. For example, infantry is not generally allowed to go into buildings unless explicitly ordered to do so. Why? Any soldiers here want to answer that one for me? So based on your short description above I am not one bit surprised that you had a unit take the "long way" instead of shortcutting through a building. It's supposed to work that way (by and large).

We're currently doing some other stuff so we haven't looked at the save files sent to JohnO, but we will soon. I can't comment since I have to get v1.08 back up and running tomorrow (and Windows people don't understand why us Mac people hate Windows ). But from his descriptions it sounds like there's a bug there of some sort. That is definitely something we want to pin down. Unfortunately, saves rarely give us a "slam dunk" view at the problem since generally whatever caused the behavior has already come and gone, leaving the effects instead of showing the cause. But we'll see!

Waypoints shouldn't be missed. However, they might look to be missed because the soldiers move to the middle of the Action Spot and the Waypoint might be on the edge. The only solution to that would be to have Waypoints "snap to grid" like Area Fire does. That's not a good solution in our opinion and in fact we have on our ToDo List removing that Area Fire behavior since it seems clear that it bothers people too much (personally, I agree).

In a general sense the TacAI is incapable of figuring out if it should hug a wall or remain spaced away from it. This is an extremely complex real world decision for a unit to make and there isn't a good way for the TacAI to figure this out on its own. It's something we would like to fix but probably can't without some sort of additional Human direction. Micromanagement options are really not a good idea, so unless we can group it in with some existing behavior it's not likely to happen.

Bottomline... CM:SF's pathfinding is not perfect, for sure. Neither was CMx1's, nor any other game that I personally have played. It's a bugger of a thing to get right and the more complex the environment, the more difficult it is to hit the unobtainable "perfect" behavior. Even with all this effort, it will never be 100% correct 100% of the time in 100% of all situations to the 100% satisfaction of the player. Having said that, when we are made aware of a problem we try to figure out what it is and fix it. We have an extremely strong record to stand on in that regard.

Steve

[Dirtweasle]

Originally posted by Battlefront.com:

Dirtweasle,

We comment about stuff like this all the time, so not sure why you think we've suddenly lost interest in this topic.

... So I guess I'm not sure why you feel this issue is just floating around without any concern about it.

I did not write nor did I make an inference that you've suddenly lost interest, or any issue was floating about without any concern.

When one of your testers wrote -

Cannot answer, NDA is kicking in!

I thought the implication was that there was an answer I might find satisfying but the tester was not able to. So I asked for BFC to answer.

Originally posted by Battlefront.com:

Sometimes pathfinding gets blamed for doing things that it should do. For example, infantry is not generally allowed to go into buildings unless explicitly ordered to do so. Why? Any soldiers here want to answer that one for me? So based on your short description above I am not one bit surprised that you had a unit take the "long way" instead of shortcutting through a building. It's supposed to work that way (by and large).

I've seen that too, and yea that makes sense.

The issue I had though was one squad went one way, and the follow on squad went the other. To my eye each squad should 've went the same way.

I've tried using many way points around buildings and through alleys in complex urban maps and found that technique very problematic due to the squad having to form up at or near each waypoint. Sometimes the squad will form up in such a way as to put a trooper into harms way. ...sometimes more than one, this also seems unpredictable to me, but I digress, the issue I was writing about is the unusual and unpredictable way squads find the waypoints.

Originally posted by Battlefront.com:

Waypoints shouldn't be missed. However, they might look to be missed because the soldiers move to the middle of the Action Spot and the Waypoint might be on the edge...

Ah... A few questions in either case. How can I see or detect the "action points" and how far apart are they? Because what happens is it is not until after I place the way point and hit "go" that it's apparent the squad in question is not going to go where I ordered, but a little ways off. Sometimes it is a few meters maybe, other times looks to me like 10 - 20 meters. Which soldier is the key? Is it the lead point man, or the squad leader? See, what also happens is the squad will to my eyes anyway break cover, and I suspect (from what you write) that this must be because the action point is past or away from the cover? Is the terrain tiling consitant with where the action points are or sometimes in conflict?

Originally posted by Battlefront.com:

In a general sense the TacAI is incapable of figuring out if it should hug a wall or remain spaced away from it. This is an extremely complex real world decision for a unit to make and there isn't a good way for the TacAI to figure this out on its own. It's something we would like to fix but probably can't without some sort of additional Human direction. Micromanagement options are really not a good idea, so unless we can group it in with some existing behavior it's not likely to happen.

Not sure how helpful this might be, but something about it reminded me of a lesson in astronomy from a long time ago how gravity and mass are related in space. Take a piece of cloth and loosely fix it over a hoop or a large ring, you know like they use for needle point. Place a dense heavy ball in the middle. Place a lighter less dense ball on the cloth and you will see the lighter ball travel over to the heavier one. If somehow you could assign a value to points along a wall or pieces of cover could you I wonder program the squad to be "attracted" to it? ...just a thought that occurred to me, no big deal if it's silly.

Originally posted by Battlefront.com:

Bottomline... CM:SF's pathfinding is not perfect, for sure. Neither was CMx1's, nor any other game that I personally have played. It's a bugger of a thing to get right and the more complex the environment, the more difficult it is to hit the unobtainable "perfect" behavior. Even with all this effort, it will never be 100% correct 100% of the time in 100% of all situations to the 100% satisfaction of the player. Having said that, when we are made aware of a problem we try to figure out what it is and fix it. We have an extremely strong record to stand on in that regard.

Steve

I have to say this continuing mantra about perfection is I am afraid getting a bit old. Nobody is trying to hold you to some perfect ideal that cannot be reached. At the very least I am not. I'm trying to understand how come the game behaves in a certain way and in the case of the waypoints defies what I see going on.

You know who Alan Kulwiki is Steve? He was a Mechanical Engineer from Wisconsin that IIRC was the last independent car owner to win a Winston Cup. Any how, Alan had a saying; "The pursuit of perfection is frustrating and a waste of time. No one is perfect. The pursuit of excellence is commendable and worthwhile. Therefore, strive for excellence and not perfection." Maybe that 'll make sense to you as it does to me.

[ChrisND]

Dirtweasle makes a good point that I think bears emphasizing. Using closely spaced and detailed waypoints can often be a real danger/hassle because the squad forms up on each one.

[ April 10, 2008, 08:08 AM: Message edited by: Normal Dude ]

[Melnibone]

I very rarely use more than one waypoint for infantry at a time. The lack of CMx1 command delays means there is less reason for me to use multiple or long waypoints. It may slow down the movement a little - but that seems no bad thing for me. It's when I rush the squads around too much that they get whacked.

[Huntarr]

For example, infantry is not generally allowed to go into buildings unless explicitly ordered to do so. Why? Any soldiers here want to answer that one for me?

You can wind up FUBAR’d, SIR!

Ahh, the age ole pathfinding issue. I think the only game to get pathfinding right was Lemmings

[Dirtweasle]

Originally posted by Melnibone:

I very rarely use more than one waypoint for infantry at a time. ...

I have found that to be, all things considered, the best bet too.

Actually to me the absence of CMx1 style delay penalties for each waypoint should mean more waypoints, and more intricate placement are more appropriate. So you could - in theory - order a squad to go down a specific alley or through a specific door, but because as it turns out the waypoint(s) adjust to the action point, (do I have that right?)and AFAIK there's no telling where that is, (is there?), it all becomes very problematic. So you end up only using as few points as you can get away with. Less variables, less unpredictable behavior.

Because the uncertainty of where the path will actually be compounds the more waypoints there are. Not because of the absence of a delay penalty.