Too painful to watch ??

[Ascendion]

Over and over I see the bot spinning the tank in place or traversing back and forth and I think I finally understand why -- the combat and navigation code are conflicting when there are nearby targets.. I just watched as my tank would take a shot, turn away because the enemy ducked behind a rock or building, then have to turn back when they popped out for another shot.... once engaged with the enemy the bot needs to stay in combat mode and either sit on previous aim point until the target dies or switch navigation to advance on the target while maintaining a bead on where the target is likely to appear.

Some access to the navigational data would make possible targeting code that is map aware and more intelligent about whether or not to release combat mode or not when a target is out of line of site.

Do I have to beg for you guys to expose more of the internal data so us experienced programmers can write nav/target code that cooperates with each other to make the bot behave more like an actual tank crew ?? I've been poking through the exposed interfaces and we already have all the visible tanks listed... so it should be possible to do nav code that avoids friendly tanks, or create a hunter/killer mode that actively goes after any sighted enemy, or heads towards a last known location to try to pick up a new sighting, etc....

 

[superreeen]

What are you trying to do should already be possible, you propably only have to implement a target provider, that's what triggers if the bot starts moving or keep firing, just take one of the provided and play around with it.

When you want to digg deeper take a look at the routines, that's the stuff that does the actual aiming.

The nav doesn't care about game state, just give it a point to nav to and it will do it's best to get there.

I hope that helps you on your mission.

 

[Ascendion]

Ya I might be able to fix the state switch from combat to roaming.. gonna look at that more in a bit... but I don't necessarily want the nav to just "go somewhere" there are times when the tactics programming I'm thinking about would need fine control like "back up 5 meters" or "back up 2 seconds" (whatever makes more sense). I notice that there is a Face method but no fine control of the turret or vehicle motion. I also noticed the GetHeight methods and a little explanation of those would REALLY help... if there were a point during scenario start up that I could scan the map using that method and build up a nav mesh with tactically useful information about sections of the mesh then it would be possible to start programming actual tactics routines to enhance behavior I need to know basic stuff like what are the 2D coord extents that GetHeight takes as input (expose a Vector2 Extents property ??), whats contained in the vector3 that the "out" version of the method creates, is the height just the terrain or does it show higher values for rocks and buildings. where can I tap into the pre-game processing to perform analysis or load a cached analysis (the earliest point where the arena is known and I can start making calls on GetHeight)... are there any multithreading restrictions on using GetHeight ??