Annoying pathfinding issue still in 1.03

[mazex]

OK, I waited before whining about this one as there where promises of clearing up the pathfinding issues in 1.03. This one really annoys me even though it most of the time does not kill my uits but it's a massive immersion killer for me. When I start up a map like "Allahs Fist" for example and order my M1:s to move between the palm trees (a course that is something like 30 degrees NE). Immediately after they start moving, they stray far to the right of the path I assigned to them in a course that seems to be more like 45 degrees. This seems to continue until arriving to the point where they can go straight to the north (0 degrees) to get to the destination I plotted for them... At this point they will have to travel THROUGH the palm trees to the right of the gap I wanted them to use. The path they choose (disobeying a clear and simple order from their commander!) looks more like the path units would have to take on a hex based game when trying to go 30 degrees.

The whole thing is getting very annoying as plotting a path in an almost flat desert should not be that hard - and there is NO reason to NOT take the shortest path to the destination that I have assigned to them (following the GD path!). This whole thing distracts me a lot as the feeling of control of their movements is completely lost. They end up among the trees trying to zig zag between them and I sit there swearing.

Apart from that 1.03 feels like a good step forward, but the FPS seems even worse on my 8800GTX than before on maps like "Allahs Fist" (I know that it's not in the list of fixes until 1.04)... I've been forced to go down to "Improved" in my settings for textures and models instead of "Better" that I used before to be able to play the game.

Regards /Mazex

[Exel]

The easiest, and in fact the best solution to the pathfinding problems would be to have the vehicles and squads follow their assigned path to the point instead of plotting their own path. That's if there's no obstacles on the path that require evasion.

Let the pathfinding algorithm work its magic when you set a waypoint without much thought about the route (without regard to obstacles in the way), ie. for RT mode, but when there's no obstacles on the assigned path, follow that.

[Thomm]

mazex points out a standard problem of pathfinding.

Compare: Toward More Realistic Pathfinding by Marco Pinter

Selective quote from:

http://www.gamasutra.com/features/20010314/pinter_01.htm

(requires free registration)

[...] Unfortunately, this simplistic solution is not very effective, because all turns are still at 45-degree angles, which causes the movement to continue to look rather unrealistic. In addition, the 45-degree-angle turns often cause paths to be much longer than they have to be [...] The actual desired path is that shown in Figure 2c, which takes the most direct route, regardless of the angle. In order to achieve this effect, we introduce a simple smoothing algorithm which takes place after the standard A* algorithm has completed its path. [...] The smoothing algorithm simply checks from waypoint to waypoint along the path, trying to eliminate intermediate waypoints when possible. [...] This simple smoothing algorithm is similar to "line of sight" smoothing, in which all waypoints are progressively skipped until the last one that can be "seen" from the current position.

But of course Charles knows this - it worked in CMx1, after all?! So there must be a different reason for the 45 degree affinity.

More on pathfinding:

http://www.gameai.com/pathfinding.html

Best regards,

Thomm

[hoolaman]

Originally posted by mazex:

OK, I waited before whining about this one as there where promises of clearing up the pathfinding issues in 1.03. This one really annoys me even though it most of the time does not kill my uits but it's a massive immersion killer for me. When I start up a map like "Allahs Fist" for example and order my M1:s to move between the palm trees (a course that is something like 30 degrees NE). Immediately after they start moving, they stray far to the right of the path I assigned to them in a course that seems to be more like 45 degrees. This seems to continue until arriving to the point where they can go straight to the north (0 degrees) to get to the destination I plotted for them... At this point they will have to travel THROUGH the palm trees to the right of the gap I wanted them to use. The path they choose (disobeying a clear and simple order from their commander!) looks more like the path units would have to take on a hex based game when trying to go 30 degrees.

Yep, after all the talking about how complaining about CMSF is like wishing hexes were in CMBO, it seems that hexes are back!

The action spots seem to be just hexes (or squares more accurately). Units tend to move like a knight in chess, 1 across then two forward. Spotting seems to be similarly effected. cf shooting through walls.

[Darkmath]

The developpers could not stay with the same 1 metres grid location engine from CMX1 ; it would cost too much CPU cycle so the game could not be played in real time.

Therefore, they choose the action spot design for CMX2.

BTW , do you remember the shooting through edges of buildings with tanks in CMX1 days?

Even LOS calculation wasn't perfect in CMX1.

Still, I agree it would be clearer for the player to see its units follow a straight line when a single waypoint is ordered.

[Pandur]

because all turns are still at 45-degree angles, which causes the movement to continue to look rather unrealistic.

that is why one plots 2 points per corner making it two 22,5° turns wich are taken in more or less one curve.

it works great in CMx1 compared to CMx2, but that doesnt mean it cant be that good at later stages...

[Mr Byte]

I plotted virtually a straight line path for four strykers the rear most started to turn immediately toward a house then reversed then forward then reverse etc while the rest moved off eventually it got back to pointing in the right direction but had a minutes gap between it and the others......

[Exel]

Originally posted by Rollstoy:

Compare: Toward More Realistic Pathfinding by Marco Pinter

The smoothing algorithm simply checks from waypoint to waypoint along the path, trying to eliminate intermediate waypoints when possible. [...] This simple smoothing algorithm is similar to "line of sight" smoothing, in which all waypoints are progressively skipped until the last one that can be "seen" from the current position.

What?! I don't want the game to eliminate or "progressively skip" waypoints I have assigned to it at its whim. If I wanted an intermediate waypoint to be skipped, I wouldn't place that waypoint in the first place!

Or is it referring to the hidden waypoits the computer plots for itself to accomplish a given route? To that I'd say; Don't. If I assign a carefully planned waypoint route, there's no reason for the computer to plot a course of its own unless there are obstacles in the way.

[Exel]

Originally posted by Darkmath:

The developpers could not stay with the same 1 metres grid location engine from CMX1 ; it would cost too much CPU cycle so the game could not be played in real time.

Therefore, they choose the action spot design for CMX2.

So another compromise for RT then? It starts to seem like that most of the things wrong about CMSF are directly or indirectly due to the RT mode.

[thewood]

I am still absolutely confused by the 8m grid thing.

So in CM1 we had very abstracted squads, 20m terrain grids, and 1m positioning grid for units

In CMSF, we have 1:1 squad member representation, 8m position grid, and what size terrain grid, if there even is one.

Is that correct?

If that is correct, with all we gain tracking individualsoldiers, is it a step back tracking to 8M, and not 1m?

[Redwolf]

Originally posted by Darkmath:

BTW , do you remember the shooting through edges of buildings with tanks in CMX1 days?

Even LOS calculation wasn't perfect in CMX1.

Uh? That wasn't the case. It was not a LOS problem, it was a timing problem.

What happened is that CMx1 decides whether a tank is hit or not at the time that the shot is fired (executing the hit probability and throwing a dice). If it had decided that the shot hits, the shot will hit. Of course it evaluates whether there is LOS. If there is no LOS there's no shot and hence no hit.

However, since the hit-or-hiss decision was made at shot time, you would then end up shooting through a LOS-blocking obstacle if the tank as moving and it was visible at shot time but out of LOS by the time that the shell travel time had passed. In the latter case you would get a shot through LOS-blocking obstacles (because at the time the shot started it wasn't blocked and that's when CM decides it will hit).

[Darkmath]

Let suppose a unit should go from point A to point B.

If the game treat the positionning, hence the pathfinding , as hexes or square, the unit have under certain circumstances have to move like a knight on a chess game : moving in a L shaped path.

There are several L path the unit could take to go to B.

It may be better for the unit to go on the side THEN go straight ahead to B, otherwise it would expose it flanks when turning to B once the units is at the same level.

[Pandur]

well, it was also possible becouse you couuld see through "corners" of houses wich was perfectly predictable and measurable with the LOS tool.

also, this counted for the "houses/shacks" only! not for every pice of terrain as we have it now(seeing and shooting through solid ground).

[Darkmath]

I played Allah's fist scenario and I ordered to a tank 1 platoon a fast command.

I also notice that the tank won't go in a straight line.

But the distance travelled along the path is too long to be explained by a action spot issue.

The same behaviour occurs for vehicles (just tried with M1) in other scenarios in open terrain.

I don't know why the hell they won't go just in a straight line. It is annoying indeed.

[ September 02, 2007, 07:17 AM: Message edited by: Darkmath ]

[mazex]

I sure is annoying, and if they follow the 8x8 grid in a "hexbased" manner it would be better to show those squares. Cards on the table gents! It would be less annoying for me at least... I don't really get this 1:1 representation of soldiers and bullets when the "real" 3D world feels further away than it did in CMx1. The underlying grid must be available for debugging purposes - give us a checkbox or a key-combo to reveal it!

/Mazex

[Exel]

If movement and terrain is that much abstracted, then I honestly see no advantage in the 1:1 representation of units. If you can't make the terrain be 1:1 with the units, then abstract the units too. This is just too confusing for the player and seriously hurt gameplay - if you have 1:1 unit models then you expect them to act like that too, not abstracted.

[thewood]

If you go back to the long(-winded?) thread that dorosh threw out right after release, I think the arguements were basically that. I have come to the conclusion, after much trying, I just don't get it.

[Cpl Steiner]

To my mind I don't care what size underlying grid there is - the game should use basic high school trigonometry to determine what bearing in degrees a unit needs to be heading in to get to a particular square. It should not just move from square to square like some demented chess piece. That sort of programming should have gone the same way as "X-COM". By all means use the underlying grid for spotting purposes etc. but that should be all.

I also don't buy the idea that the game should second guess me and decide it needs to keep a certain facing towards the enemy rather than just go in a straight line. I have the option to give it multiple waypoints for exactly that purpose.

I am hoping that these anomolies are not down to design but are simple bugs that we will eventually see the back of.

[Exel]

Originally posted by Cpl Steiner:

I am hoping that these anomolies are not down to design but are simple bugs that we will eventually see the back of.

Or that it would at least be acknowledged that there is a flaw, whether by design or not, and action would be taken to correct it - improve the design if it's a design issue, fix the glitch if it's not.

[hoolaman]

If this all worked like it was originally described, I don't suppose there would be an issue.

Steve originally stated that the action points (which I assume to be 8x8m tiles), were only used for calculating LOS. This was meant to be an abstraction that wouldn't be too noticable but would result in occasional odd results. As it stands the shooting through walls seems to be an unintentional result of this system.

LOF, movement and I suppose nearly everything else was supposed to have been resolved much more finely presumably down to less than 1x1m for individual bullets.

What we seem to see ATM is the spotting tiles used for movement (one right two forward), area fire (snap to grid), and producing odd results with squads repositioning themselves in the centre of a square.

[Darkmath]

I read LOF is resolved down to less than 1*1m , but I didn't know the movement is stated to be in this resolution too.

[WW2Wargamer]

Hi

IMHO Pathfinding, Los, and most things got better from patch 102 to patch 103. As i stated before battlefront WILL FIX most problems. This is a game that cost $50. Most of us will play CMSF for a long time. Which means REPLAY VALUE! Give Battlefront a break, how many game companies give you the support they do! They listen and tell us what they can and cant fix. Its terrible their HONEST. Think about it REPLAY VALUE, BIWEEKLY PATCHES, HONESTY, and a game that gets better and better.

Thanks and enjoy CMSF

[Exel]

Originally posted by WW2Wargamer:

As i stated before battlefront WILL FIX most problems. This is a game that cost $50. Most of us will play CMSF for a long time. Which means REPLAY VALUE!

Right now I'm rather looking for the play value, which atm isn't really there with all the issues the game has. Knowing Battlefront I'm confident they will be keep supporting the game for a long time. But that doesn't mean we should stay silent in our confidence - if the issues weren't brought to light, how could they ever be fixed? And it's not just what gets fixed and how, but also when - priorities of the patches are something we also need to influence, so that the most critical issues get solved first and not after 2 years.

[SlowMotion]

...and also: the more bugs are fixed now, the better things should be when the CMx2 WW2 game arrives.

[WW2Wargamer]

Hi Exel/SlowMotion

I never said not to complain. But when the complaints are the same things over, and over it gets to be a pain. The game IMHO is getting better! I also think that we should let the new people know that things are getting better. Its

counter productive to our Hobby to keep talking only about the negative.

Thanks