This is not a bug report thread, but more of a discussion thread about current pathfinding bugs, glitches and general weirdness, and possibly even talk of when it's working properly! For example, this is your typical 'cost field' info, with some stuck units as a bonus: These stuck units are actually close to the end of a very long move order they were given. They will probably free themselves if given new orders, which might be an interesting cause for speculation. If the chasm didn't have nooks like the one where the bots are 'trapped', Doom era jostling would probably get the bots around it anyway. (Also, towards the top, you can see the grid by the metal extractors is getting skewed and chopped. I wonder how that ultimately affects pathing?) Still, what if there's something else at play? This is using a different mode of the navigation info display: The bots on the left are going to be ordered to move to the bots on the right. If the order is placed on the pink side of the planet, the bots will take the path around the lower half of the screen. However, if the order is placed on the green side of the planet, the bots will walk the long way around the T2 factory and the top half of the screen (essentially they'll refuse to walk on the pink sector if their orders don't require them to, even if the path is shorter). Really, I wish I'd taken screenshots of this stuff while messing around in the alpha.
I brought up this idea a bit ago, but... The path finding foot print (IE: the red and green blocks around the base of buildings and other stuff) should be increased for everything by about 25 to 40 percent to make certain it sticks out from the object. The path finding algorithm is fine. It is just being fed bad data. The physics collisions and the path finding are working against each other because the foot print is not big enough in most cases. The green blocks in the first pic should never even come close to touching the edge of the crack and in fact should be red (unpathable) and the entire footprint expanded one square.
New stuff. Metal extractors. The one on the left was placed second, and had to be rotated before it would take. The blue bounding boxes create curious shapes in the pathing grid, and are probably the actual measure that buildings use to determine the validity of placement. (Something to consider for UI design later?) While the piled up tanks are caught on a slight pathing nook, you can see a line of other tanks at the top, part of the same group, that were simply pushing against the obstacle and sliding off by sheer luck. Apparently that order didn't invoke any obstacle avoidance calculations in the units. The group of tanks on the bottom were given a separate order and to their merit they are properly avoiding obstacles.
Oooooh dem metal planets... my units get stuck there like the pages of a nudie magazine stashed under a young boys bed
Well, they did just do some work with it. But I won't be able to be near a broadband connection for a the next few days.
One thing I'm wondering is how far can units perform a lookahead when pathing? It certainly doesn't seem to be very far, and I wonder what the cost would be to temporarily increasing it to at least the units actual vision range. For example, in the first picture the units can clearly "see" how to get around, but the pathing routine can't seem to use this data. Also I wonder if it's possible for the units to sort of network, and stuck units query other units in their group on how they got around the obstructions.
Flow fields are kind of strange so I'm not sure if it's even possible to cache partial paths. I mean there are are no nodes so what data would you cache from the path in this case? The lack of nodes may also relate to why units won't go across a map boundary even if crossing the boundary would result in a shorter route. How do you measure distance with a flow field (sanely)?
