Regarding the path nodes of Bots (on the map “Islands”)

[limsup]

In a recent match played on the map “Islands”[standard edt.], the number of human players being small, so 20+ A.I. players were voted in, to “fill the gap” until more people join. When I joined in, I found it curious, that 4-5 bots were standing still on the interior middle ramp of the Hand-of-Nod. At first I dismissed it as a temporary glitch in the bots path-finding routines, but shockingly soon enough other bots have just kept running in to that “central mass of bots”, and got caught up in it! At the end of things, some 5-7 minutes after my first observation, a fellow player counted 17 bots standing in a roughly 4 m² area without moving a muscle! (see screenshots)

Suffice to say, the enemy team gladly capitalized on this, conducing the collapse of our defenses. This has led to the Nod team’s undoing, as we could not terminate the affected bots, nor could we relocate them elsewhere, while the GDI bots came pouring in our base.

Thus, I would like to ask for the kind opinions of the senior colleagues on how should such a situation be handled in the future, and what could we, “ordinary” players do to resolve such a phenomenon, in situ, or to avoid it altogether? Considering this could adversely affect a team, and the opposing team would most likely veto any public voting aimed to resolve such a situation (as war is about opportunism), what could you suggest?

Lastly, I would be interested in the underlying cause of this phenomenon. Is it simply a way-point related issue, or could other factors play a role in the accumulation of bots in a single point(nothing obstructed their way, except they themselves)?

Should by chance Fffreak9999 read this post, his opinion would be very welcomed J (I think highly of his in-depth answers)

[Fffreak9999]

As a quick question, was this on Islands or Islands_Flying?

[limsup]

Islands, baseline version, not Islands_Flying...

[Fffreak9999]

I would guess it has something to do with the single node there, possibly a Destination only node.

 

I would need to look to verify.

[ShrewdTactician]

I've been working heavily on the bot navigation (and bots in general), mainly looking to ditch the old waypoint system for the most part and replace it with the more modern navmesh system. Strangely I have been using islands as my test map. Navmesh have numerous benefits over the waypoint system, and should allow us to do more with bots overall. This is all "experimental" and could turn out that if I cant get it to 100%, then it will all have to be dropped, but of the work I have done so far, its looking pretty promising. I'm actually quite happy with some of the improvements to the bots already.

[Madkill40]

@ShrewdTactician Will Harvester pathfinding also be improved?

[ShrewdTactician]

Should be. Instead of following waypoints, the harvester finds it's own way using the navmesh, which is a precomputed map of movable space. The navmesh is devided into sections, with the edges of these sections marked with the size of actors that can fit through it, including height. 

So harvesters shouldn't run into things, doesn't rely on map makers having to place waypoints down everywhere, the bots have alot more freedom in how they move, and where they can move. 

[Madkill40]

So how does the navmesh work in regards to map making? Volumes over the terrain? Some crazy cool numbers game?

[ShrewdTactician]

I am planning to make a tutorial video for map makers, as well as converting all published maps that could he used for examples. I wanted to make it as easy as possible for mapmakers,and should be easier then the current waypoint system.

For a quick run down. You put down a navigtion point called a pylon, and use either a radius or volume(s). Cover the areas in your map that are accessable and then press build paths. Done.

The Udk "explores" the pylon area from the pylon, creating a "map" of all the areas that can be reached, this includes drops off edges, ledges, jumps,mantling, etc(which are now aware to the bots). 

The bots now use this map to work out all the path finding. Since it deals with areas rather then points, the movement is more free and "natural". It also makes pathfinding cheaper, as the map is easier to check against then trying to find the nearest waypoint node of the source and destination of a trip, which has to be line checked multiple times.

The other advantage for map makes is, that except for major region changes, you won't have to keep muking with pathong when stuff changes. The amount of waypoints I've found underground, in building walls, etc is nuts. That hassle for map makers should go away.

[limsup]

4 hours ago, ShrewdTactician said:

I've been working heavily on the bot navigation (and bots in general), mainly looking to ditch the old waypoint system for the most part and replace it with the more modern navmesh system. Strangely I have been using islands as my test map. Navmesh have numerous benefits over the waypoint system, and should allow us to do more with bots overall. This is all "experimental" and could turn out that if I cant get it to 100%, then it will all have to be dropped, but of the work I have done so far, its looking pretty promising. I'm actually quite happy with some of the improvements to the bots already.

Thank you for the valuable insight!

I think that any advancement would be most welcome! If this could be realized, it could mean a huge leap forward in terms of game-play, especially regarding the harvesters, which tend to become stuck on a number of maps...

[Madkill40]

You're right, placing over 1,000 pathnodes can be a bit of a pain especially when they go missing.

I wonder if the change to a navmesh will prevent bots from suiciding as often as they currently do.

[ShrewdTactician]

I do want to fix that aswell. Sometimes it's them getting stuck that causes them to suicide but also sometimes it's just their logic that causes it.

There's essentially going to be "parts" to the bot upgrade. The navmesh is part 1, which deals which navigation/movement/freedom. Part 2 is the "logic" part, where I'm going to teach them how to play renegade. Teach them to target the mct, how to eng rush, or tank rush, how to use stealth, etc. Part 2 might get split into a third section where I move the logic out of hardcoding and into the visual scripting kismet. 

[RoundShades]

Sounds promising. I sort of wish you were on the development team for Monday Night Combat. (if they ever start up a new game or sequel, I'm going to recommend you send an application, and they look into it.

[limsup]

5 hours ago, ShrewdTactician said:

2 is the "logic" part, where I'm going to teach them how to play renegade.

A* search, ANN with Newton- and least squares methods?

[ShrewdTactician]

I'm not sure if you quoted the right line. 

The navmesh(part 1) uses an a* search for the edge to edge pathbuild on the polygons in the navmesh. 

 

[limsup]

Sorry, my mouse was too sensitive