Matrix Games Forums

ORIGINAL: Dundradal

I think it has to do with how the vehicles are comprehending the bridges.

That's exactly it.

Steve had made an explanation in another thread about this and that was one component of his post. How the pathfinder perceives "better" is the key here. Many factors at work but depending upon the movement type, ease of egress, known threats, cover on all the possible open routes, etc. going around the bridge might be "better" in the eyes of the pathfinder even though we disagree with it. That can be minimized. That will always be a component of the pathfinder as it's impossible to get it to make decisions we agree with 100% of the time. Not saying it can't be improved it's just that it's important to make the distinction.

*EDIT* And to the larger point you were making... thanks for the extra info re: waypoints vs. no waypoints. Every bit of info like that helps.

I also noticed it had to do with distance to the bridge. The further away from the bridge the more likely the vehicle would either give you the "can't go there" command or it would take a unique route that made little sense. The last sherman to cross did so because it decided to wander around the bottom of the map. I gave it a move fast command up the road but it decided to go through the woods.

And I've done some testing myself so I know that every factor is important.

In another thread when someone mentioned how the tank went through the woods besides the bridge, I experienced the same thing. I wonder if perhaps the terrain is coded wrong and the tank is actually trying to cross where it thinks the bridge is?

Thanks again for the extra info. I remember you telling me you'd done some testing. I could see your experience at work here. [;)]

Before release we'd gone through the maps several times making sure the coding was correct. Can't hurt to check it again but we're fairly certain it's related to the pathfinder as this map is not the only one having an issue. Good thought though. Thanks!

Bridge looks coded fine

Elevation shows the humped bridge

might be worth trying with the same elevation as the road?

why the coder painted nipples on the map I don't know [X(]

Hehehe and that is weird. Something is causing the tanks to want to cross to the left.

Not a problem Oddball. I'm just not used to how CC works as I've only recently gotten back into the game after years away.

Just tried it with changed Elevation to match surrounding area.... no dice.

Next trick... remove the bridge code and replace with " Paved Road"

nope, that didn't work


remove the Bridge walls?

yep, that's it... remove the bridge walls and voila.

Now to put back the bridge and the raised elevation but no bridge walls just to prove the theory (tested)

Done... works like it should...except being able to drive off the bridge sides.

So the bridge walls are the issue?

it sure looks that way... it is how the exe deals with the "GAP" I believe.

Checked the vehicles.txt and almost all vehicles have an element radius of 2. Thus the 6 wide gap/opening between the bridge walls, that are impassable, on the southern end of the bridge should be more than enough.

Good find Andrew, hopefully that will help the programmer resolve the issue....

Sorry to add more issues related to pathing but not doing so will never help.

My worry after that last few posts; please don't just focus on that single bridge and think it is solved, that vehicle pathfinding is difficult everywhere.

I encountered problems with pathfinding from the first vehicle to the last having played the GC for two days. So that's only the map South of Eindhoven and Eindhoven actually. Although touted to have been improved, the pathfinding actually seems worse than before. Main issues:
1. Long routes (25+ meters) are almost always a sorry affair.
2. Tight conditions (roads, bridges, build up areas, etc.) get that "Can't go there!" result very often.
3. Twisting and turning before starting a move.

Ad. 1. The long routes.
Even in very straightforward conditions e.g. the long wide main road in Eindhoven, the vehicles bounce from left to right. It is as if looking for the edges to determine passable terrain. Could that be the problem?
Ad. 2. Tight conditions
But even when ordered to move only a few meters, say only 10, you see a lot of route checking and "Can't go there!" messages even when it's utterly clear it is possible. This even happens on the long clear road to Eindhoven but gets very much worse when in tight conditions. Almost like every little map feature means a possible road block. In Eindhoven for example it was as if the tram rails in the road or the road sides were routing problems.
Ad. 3. Starting a move.
Even the most basic short move almost always starts with the vehicle turning left and right before setting of. That's also the case when ordering a short move in the same direction.

Using vehicles is a very frustrating affair currently. More so than in former CC games I think all of which I played and all the mods. What's changed in this version of path finding compared to the former games? It didn't work and might be the cause of the matter . . .

Only viewing the current wayward routing as command problems helps me playing the game which is still exciting.

Good luck bug hunting.

Andrew, nice detective work! It certainly seems that the way pathfinding is calculated might need to be fundamentally changed after reading your results and Jaroen's below.

Jaroen has hit many of the issues on the head. I've noticed all the same things at times. I've seen this "pathfinding in XX game is the worst of the lot!" and I think I've got a theory. With each game I'm guessing the complexity of maps and perhaps some other factors are increasing. I'm going to wager that the pathfinding routine has been improved, however in the process of improving one thing it may have broken another in some way. By the number of "can't move there" and just wacky paths vehicles take it seems that the routine used to find a path is being overwhelmed and is "breaking" when the path gets too complicated.

Removing the bridge walls is removing the obstacle completely.  You basically just have open terrain.  There is no obstacle to navigate in this case.

The underlying problem has to do with the implementation of the path finding system from the original Close Combat.  There are several issues, but it is mainly how the high level path finding abstracts the terrain and weights possible paths -- this is done to find a path quickly, in real time.

And, as Oddball said, we are discussing how best to address the issue.

Steve

Not a problem...just here to help add input to your discussion. [:)]

It's many, many moons since I last toyed with the innards of a CC game 'base' folder, but I was wondering... Has anyone ever tried to improve pathfinding by stripping complexity out of the 'Elements' file?

What would happen if, say, instead of 20(?) different movement modifier values in the Tracked column there was three? A 'fast' value for road-like surfaces, a 'medium' value for field-like surfaces, and a 'slow' one for rubble/streams etc. Something similar would need to be done with the bog chances. There'd be a loss of granularity, sure,  but I wonder if a) it would be all that noticeable and b) the simplification would help the pathfinding routines make more sensible decisions.

Just an idea. Sadly, I'm not in a position to test it at the moment. Do the modern CCs even have an elements file anymore?

 

ORIGINAL: Andrew Williams

Amazing discusion and regardless of the outcome, I say you've discovered a graphics look that could just be the way to get the younger generations interested in war gaming