Unit Movement on 2d Grid

This topic is 1649 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

Recommended Posts

Heyho,

I am designing a grid based game, well 2-d, flat ... the grid is primary for the astar path finding algorithm.  So i got to the part where i'm thinking how the units are going to move.  My initial thoughts were this "unit step size" .. every server game tick a unit will try to move its full step size, I think of it as being the speed of the unit, the higher the step size, the greater the speed of the unit.

Anyway as it panned out, i came into some problems, since i have a path which a unit must follow (a list of waypoints essentially) .. i figured i need to make sure that the unit always moves the full unit step size (unit.speed) in each game tick.   Okay so the problem arises when the pathing changes direction .. Example: 4 up, 2 right ... when it turns right ... .

I would divide the (4 up *grid.size) by unit step size (unit.speed).  Some occasions(mostly) unit step size does not fit into that distance wholly.. so there is left over( remainder ) ... same like 20 / 6 = 3 remainder 2.   Question is .. what to do with this remainder?  of course im trying to find solutions but now its made me doubt the entire idea of moving a unit to an absolute position ( with floating point precision ) , and i'm wondering if i should be moving units only to fixed rigid points of a grid?

I have played games like warcraft 3, i made a test map in there , got it to print my position and i noticed that it seems to be using absolute positioning, if so, how has it done it ? :( .. i liked the idea of absolute positioning becaues i felt it would allow more control for the user .. but then realised that a very small/granular grid would be almost the same and un-noticable for the player.

help is much appreciated.! i've been searching for some guys talking of similar thing on internet for hours without any success. chhers!

Share on other sites
so your a* does not organically allow for diaginal movement and you are trying to reconcile this?

Share on other sites

I have been thinking about a solution and the idea i've come up with is to have the server send 2 sets of data. One set is for the position of a unit , and the other is intermediate points which correspond to changes in direction.  I decided that these changes of direction are important for the client to know, so that it can make the unit move to the point which a unit changes direction during interpolatjion .. when 2 points are outside of that change of direction point.

@Orynmus the astar does not no , I think i will filter the astar points and skip certain tiles if there is no collision between 2 , to make shorter paths

Share on other sites

So, if I am interpreting this correctly, you have each unit moving  some distance, but sometimes the distance is greater than the distance between two waypoints?  It sounds like the client needs to know the path, so that when it gets information back from the server it can interpolate the unit along the path.  It probably needs to know that anyway, as many games benefit from showing the path to the user.

(or you need to drastically reduce your step size, or cull waypoints that are smaller than the stepsize, though that might have units moving through blocked tiles.)

Edited by ferrous

Share on other sites

In my experience, A* solutions frequently boil down to two-part answers.  You define a grid with just enough fidelity to accurately depict all obstacles, and use A* to navigate that grid.  Once you've entered a cell, that cell is now blocked and other obstacles or actors aren't allowed to move into it until you vacate the location.  Once you've gone as far along the grid as possible (you're in the target cell or as close to the target cell as possible), you navigate directly to a specific destination within that cell.  You don't need to consider other obstacles anymore, because the definition of your system prevents other obstacles from consuming the cell while you are already inside it.

Given your description of your architecture, I would expect the server to continue holding onto the computations for all full-cell movements... but once inside the cell simply allow the client to finalize the position on its own.  The server doesn't need to know anything more than "I am occupying this cell."

Edited by Thinias

Share on other sites
Is your world really a grid?

Most games I've worked on used navigation meshes rather than regular grids. The modelers who build the world build a second mesh that overlaps the graphical ground but is used for navigation. It may be used for physics as well, or a third mesh may be used for that.

With a nav mesh, even if your nav mesh is 2d, you can still have your arbitrary precision. Navigation nodes are the nav mesh rather than a regular grid. The world building people can be instructed to make nav mesh regions as large or small as you want. You can also add build steps to ensure a minimum size or a maximum size, because everyone makes mistakes.

You might choose to lock others from the area as Thinias wrote above, or you can run a second collision pass against anyone else already in that nav mesh node. Remember that the big reason to use these is to reduce the problem space. Once you have the nav mesh or similar spatial tree in place the pruning is done. Almost always the node will be empty except for that actor. Very rarely there will be another actor in the space to deal with, or possibly two or three if the world builders made large nodes. Either way, at that point the problem has been pruned to such a small number that solving is not compute intensive.

• What is your GameDev Story?

In 2019 we are celebrating 20 years of GameDev.net! Share your GameDev Story with us.

• 13
• 9
• 15
• 14
• 46
• Forum Statistics

• Total Topics
634056
• Total Posts
3015282
×