Funny Map survey

[moonsense715]

Is that RA_Bunkers?

[MPRA2]

You know what? Why dont we just remove defenses entirely so players need to rely on skill and THE MOST ALIEN AND MYSTERIOUS WORD KNOW TO ALL MANKIND!!!! "TEAMWORK" to win... No AI defenses to save your ass here folks. Oh and, remove barracks and WF because they are OP.

Edited June 11, 2016 by MPRA2

[TeamWolf]

Is that RA_Bunkers?

 

Nope, Stormy Valley!

[MPRA2]

 

Is that RA_Bunkers?

 

Nope, Stormy Valley!

 

will soon be called gloomEMO_Valley

[Raptor29aa]

 

 

Is that RA_Bunkers?

Nope, Stormy Valley!

will soon be called gloomEMO_Valley

Not enough angst for it to be emo, but add a cathedral and it could be goth valley.

Anyways it is good to have some in base lighting! Also with a few infantry sweet spots on the hills, for some good rocket soldier tom foolery

[MPRA2]

I still think we should just remove defenses and replace them with ore silos or walls....

[Isaac The Madd]

Nah, just replace them with nuke launchers, and replace the free supply truck with a demo truck.

[Einstein]

I'll hold to the nuke sniper

 

Also yes, hooray for in-base lighting Can't wait to play this!

[Pushwall]

Fissure rage

 

 

[Raap]

The AI is too poor to see widespread implementation, plus it will be a consistency issue.

 

Also, one doesn't just add AI logic to a map. You need to design their functionality (their stats as well as their spawn logic such as WHERE do they spawn and how many of them spawn under what conditions, and depending on a structure being functional or not?) as well as implement pathfinding on each map, which is no simply task because the pathfinding logic doesn't understand various concept such as water, and at the same time there is no way to create an "AI world" mesh in 3DS Max since no collision settings can be set up in LevelEdit to be pathfind-collision-only. Temporarily objects can be used but is just messy.

 

And let's not forget the elephant in the room of the AI being unable to attack buildings unless a fancy programmer dude can make building mesh register as valid targets at their origin position (which even then would mean changes to some buildings as well, most likely). (Edit: Yes, building meshes, not buildings as grouped meshes with a single origin target point, that'd look weird.)

 

You really can't decide to randomly just add some bots to some maps. Doing random and poorly thought out things is what got you Gamma.

Edited June 14, 2016 by Raap

[moonsense715]

The AI is too poor to see widespread implementation, plus it will be a consistency issue.

 

Also, one doesn't just add AI logic to a map. You need to design their functionality (their stats as well as their spawn logic such as WHERE do they spawn and how many of them spawn under what conditions, and depending on a structure being functional or not?) as well as implement pathfinding on each map, which is no simply task because the pathfinding logic doesn't understand various concept such as water, and at the same time there is no way to create an "AI world" mesh in 3DS Max since no collision settings can be set up in LevelEdit to be pathfind-collision-only. Temporarily objects can be used but is just messy.

 

And let's not forget the elephant in the room of the AI being unable to attack buildings unless a fancy programmer dude can make building mesh register as valid targets at their origin position (which even then would mean changes to some buildings as well, most likely). (Edit: Yes, building meshes, not buildings as grouped meshes with a single origin target point, that'd look weird.)

 

You really can't decide to randomly just add some bots to some maps. Doing random and poorly thought out things is what got you Gamma.

I think you have an overcomplicated theory about AI implementation. I did several tests that lets me spawn and remove them on any map easily based on whatever event I wish.

[Pushwall]

Problem with Fissure bots: You can't alter the pathfind grid during the game. So with the introduction of that new destructible platform, either bots will have to completely ignore the middle lower route, which will make it really easy for players to destroy the enemy barracks, or they will be able to go through the lower middle route, which will mean when the platform gets destroyed, bots will end up piling up against the rubble that they will try to run through but can't. And almost all of the bots will do that too since bots prefer to take the shortest possible path to their enemy - which would be the lower middle path.

 

Problem with Pacific Threat bots: I already have something to help with the silo problem. Plus getting bots to "guard" something involves artificially constricting the pathfind grid because they are too bloodthirsty to stay near an object. And since they refuse to shoot infantry that are standing in a place that bots can't move... you do the math on that one. The alternative is just making them not move, but that makes them really easy targets for any weapon, especially the thief's Silenced Beretta with its nonexistent spread.

[moonsense715]

That's a problem with a pathfind system, which is another big topic and yes the pathfind sucks.

[Pushwall]

I should mention that one of the other maps will have some bots in the next version. They suffer pathblindness too, but luring them out to the edge of their pathfind grid is probably not the best idea - they're neutral bots, so that just means the other team can more easily reach what they're guarding.

[Killing_You]

The alternative is just making them not move, but that makes them really easy targets for any weapon, especially the thief's Silenced Beretta with its nonexistent spread.

 

Plus this just makes them a diet base defense, which is kinda pointless.

[Raap]

 

The AI is too poor to see widespread implementation, plus it will be a consistency issue.

 

Also, one doesn't just add AI logic to a map. You need to design their functionality (their stats as well as their spawn logic such as WHERE do they spawn and how many of them spawn under what conditions, and depending on a structure being functional or not?) as well as implement pathfinding on each map, which is no simply task because the pathfinding logic doesn't understand various concept such as water, and at the same time there is no way to create an "AI world" mesh in 3DS Max since no collision settings can be set up in LevelEdit to be pathfind-collision-only. Temporarily objects can be used but is just messy.

 

And let's not forget the elephant in the room of the AI being unable to attack buildings unless a fancy programmer dude can make building mesh register as valid targets at their origin position (which even then would mean changes to some buildings as well, most likely). (Edit: Yes, building meshes, not buildings as grouped meshes with a single origin target point, that'd look weird.)

 

You really can't decide to randomly just add some bots to some maps. Doing random and poorly thought out things is what got you Gamma.

I think you have an overcomplicated theory about AI implementation. I did several tests that lets me spawn and remove them on any map easily based on whatever event I wish.

 

 

In my opinion, I'm not over-complicating it. You need a structured design to follow as guideline for all levels so that the implementation is consistent across all levels. Map-specific mechanics aside of course, but I assumed the request was for AI presence in all maps.

 

I'd personally only design AI for all levels if it only kicked in when player counts are below 3 or 4 per team. And that would still mean you need to design what their purpose is besides simply existing. Not to mention that sending AI to objectives via waypoints and scripts would need a consistent implementation approach - but this is beyond my knowledge as I've never implemented AI in such a way before. Unless of course the design demands that these AI units are on base defender-only duty, but without changing the way the AI can identify targets beyond their generated space, this would cause issues, as Pushwall already pointed out. Creative work-around can be done on this such as creating a separate AI firing range pathfind generated region that does not connect to their travel space, but that's messy trickery and untested, not to mention a lot of extra work for each level.

 

Another lovely problem that we then ignore: generating pathfind on APB levels makes LevelEdit crash due to memory issues, because of their size. The tools were designed for box canyon maps of CamosCanyon-size. Now it is my understanding that new level tools are on the to-do list, but that's still quite a way off.

 

And icing on the cake; I'd personally want script logic that allows me to change AI behavior or send customs based on the health percentage of the AI unit. Meaning a way to make an AI unit "panic" or act more defensively once his health drops below a certain point, or start using a different weapon. I'd also like to see AI health reset to full if they remain unharmed for 2 minutes or so (essentially a modified functionality of the existing health regeneration effect), while ensuring the percentage based triggers still function additional times after such a health reset(s). And then some.

Edited June 15, 2016 by Raap

[moonsense715]

The number of AI in my implementation is gradually decreasing as players join the server, making the battle always interesting even on low player level counts. Plus removing bots from interfering games with a lot of players.

Since Saberhawk halved the memory needed that the pathfind generation uses, it might be an idea to retry it on some smaller maps that crashed before in APB.

I have a pathfind blocker around all water meshes in AR, plus a pathfind generator ON the water. Result: AI don't try to drive onto the water but they still shoot naval units and hovering units (e.g. Robot Tanks).

And the AI have no problem destroying bases either (or defending/repairing their own), since I taught them how to shoot buildings. Simple: target markers on the building walls that the AI shoot at to damage it.

The gameplay is much more interesting compared to that joke map in Renegade Skirmish mode.

 

And yes it took a ton of time to get all this working, but now it takes 3-5 minutes to add bot support to any map. Of course there is room for more improvements but it's already at a quite enjoyable level.

[Pushwall]

The issue with full pathfinding besides just possible LE crashes is that Ore Trucks don't always want to take the shortest possible path. I posted this image previously on River Raid's release - currently the Soviet OT takes the red path because the pathfind grid only extends out as far as it needs to for the OT to take that path, while with full pathfind coverage it takes the cyan path for some bizarre reason. Using pathfind blockers to funnel the OT any more than it already is will also funnel other AI units, and lord knows that front entrance will become cramped since vehicle bots don't know how to not block ore trucks. So here's another map that a bots version would be problematic for.

[moonsense715]

Have you asked the TT team why this would happen? Have you tried removing/adding more terrain objects to see how they impact route choosing? There's got to be some reason why it chooses the light-blue route but yes finding it out takes time...

[Raap]

The number of AI in my implementation is gradually decreasing as players join the server, making the battle always interesting even on low player level counts. Plus removing bots from interfering games with a lot of players.

Since Saberhawk halved the memory needed that the pathfind generation uses, it might be an idea to retry it on some smaller maps that crashed before in APB.

I have a pathfind blocker around all water meshes in AR, plus a pathfind generator ON the water. Result: AI don't try to drive onto the water but they still shoot naval units and hovering units (e.g. Robot Tanks).

And the AI have no problem destroying bases either (or defending/repairing their own), since I taught them how to shoot buildings. Simple: target markers on the building walls that the AI shoot at to damage it.

The gameplay is much more interesting compared to that joke map in Renegade Skirmish mode.

 

And yes it took a ton of time to get all this working, but now it takes 3-5 minutes to add bot support to any map. Of course there is room for more improvements but it's already at a quite enjoyable level.

 

I think you proved my point!

 

You took effort designing your AI functionality for your project, and used custom script work as well as a consistent implementation plan to ensure it functions in other maps. That's not just randomly tossing in some AI.

 

APB has nothing like that, so before APB can/should receive AI support, it first needs a design for it, which falls to Pushwall to decide as he controls the project. But I don't think W3D projects should outright copy each other, and I'd personally like to see a borderline RPG approach to AI behavior for APB, in other words, being able to trigger AI behavior/statistics/weapon changes based on their health percentage. Another example: If health drop to 25%, AI soldier finds nearest object of a certain type and crouch near it (typically anything LE-placed like tree's or dummy editor objects near rocks or buildings, all things that can be proxy'd and often already are.), said object is to be considered a cover object, said object in turn could have a script that raises nearby AI soldiers protections by X% to simulate cover). The ultimate goal is to make an AI that behaves in a unique way and eliminate the "dumbness" of it.

 

Regarding pathfinding and waypoints for Ore Trucks, I can confirm this is an ancient issue. As soon as you generate space beyond the intended Ore Truck path, they generally end up taking on a life of their own. Unexplained routes get taken, or odd maneuvers get made on the path it is supposed to take, such as awkward turns on the spot where it is supposed to be able to go into a straight line as per waypoint and terrain design.

Edited June 15, 2016 by Raap