Jump to content

brightcandle

Member
  • Content Count

    233
  • Joined

  • Last visited

  • Medals

Community Reputation

104 Excellent

About brightcandle

  • Rank
    Staff Sergeant

Contact Methods

  • Website URL
    www.charliefoxtrotops.com

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

  1. New release today for 2.0.0.alpha3. Changelog + Added getters for suppression, boost and woodland reduction so a nice 0-1 can be returned for how impacted the unit is. See the github for the syntax (or check in the functions listing for CF_BAI).
  2. I am not seeing the same problem I can connect to our server with the hotfix applied both client and server-side. So on our server the basics work.
  3. Released v2.0.0.alpha2 Fixed a bug caused by the latest game release that nerfed the suppression effect. Fixed a bug in the suppression decay which made it too rapid. Added the boost module (see the website for details)
  4. Sorry for the late reply, I missed your message somehow. When I looked into this last I found VCOM retrieved the suppression value and did things based on it but it didn't set or adjust skills based on it. So you may just find that the effect from CF_BAI is too harse for your liking and you can just tone down the impact in the settings. The default is very aggressive at 20% and it makes the AI pretty gimped. Try 50% or higher.
  5. brightcandle

    Arma 3 - Creator DLC Discussion

    The model being used here means that even those excited about this and buying it likely won't get to play the content within their existing communities and that is a real shame. I think that will limit the sales quite a bit. I notice that the Steam storefront page for the paid DLC makes no mention of the very different model under which this is being released and I think that needs to change before they start taking money for it since it is such a different proposition. They have a lot of communicating of the specifics to do before this is release ready and not a lot of time to do it or more than likely they are going to see refund requests when players work out this one is nothing like the prior DLCs with no warning.
  6. brightcandle

    Arma 3 - Creator DLC Discussion

    In terms of the assets that this paid mod contains will there be any difference in the usual EULA and terms and conditions of use? Are we free to use them to make our own missions and mods using the assets? I don't mean ripping assets just referencing their use for terrain creation and other mod purposes?
  7. brightcandle

    Arma 3 - Creator DLC Discussion

    How does this work from a dedicated server perspective. A few particular questions but a general idea of how this behaves DLC like or mod like: 1) How does the paid DLC get added to the server? Is it just there in a reduced data form or do we need to add something? 2) If playing non paid DLC terrain and content can players still join and play on the server? I am trying to work out the logistics and whether to run this we need a dedicated server specifically for the mod as not the entire community will be buying this and given that the last thing I want to do is break the primary game if it is handled more like a required mod on client and server rather than existing DLC where clients and servers can differ and still see content used by those that own it, being mostly excluded from the Tanoa terrain itself. Just wanted to check that despite this being an external parties content that BI will now be indemnifying and ensuring the DLC works despite future updates of the game. This has always been an issue with paid DLC and I wanted to make sure that such terms where in place so purchasers don't end up finding the content doesn't work in some way in the future due to game updates with no recourse.
  8. I am glad the mods are proving useful. I am trying to get some new toys out as soon in CF_BAI as we have 3 major steps to take it out of alpha. Unfortunately other issues (including the 1.90 mess and the doMove bugs) are getting in the way! We haven't published cf_weapons. I am not sure anyone else would want it. Our group focuses the game towards realistic weapon ranges, so M4's are being used at 300-400m and not 600+. Due to the way vision works in the game we end up using mostly red dots. So we end up changing a lot of firing modes to ensure all the weapons have real-world ranges/comparable ranges to a human. Amongst all that we also change a bunch of other stuff like flare lighting and adjusted dispersion values to balance weapons against each other and the appropriate ranges. It depends on ASR_AI so that it can override settings, so it has a bunch of issues with being released since it contains a real mix of stuff and is how we tweak our balance beyond just the skills. Not sure it has much value outside of our group.
  9. We replaced TPWCAS_A3 when it broke with 1.88 with the new suppression behavior in CF_BAI 2.0.alpha1. We were never really happy with it because the behaviour changes it contained interfered with ASR_AI and MCC and were very predictable. The suppression system we have made is worse fidelity and shorter range and just overall less complex but also doesn't have the behaviour impact and that does make combat a lot more dynamic. We still use ASR_AI, we did test VCOM AI recently (and I test CF_BAI with it) but it has a mix of micro and macro AI concerns and that clashes with other things where we do it better. Our AI mix is a combination of mods. We run a customized version of MCC with a lot of GAIA enhancements which does our Macro AI, CF_BAI, ASR_AI, CF_weapons changes a lot of the balance and then CF_woodland impacts the AI as well.
  10. I tested the premise of my own 7 point about whether its spotting or firing and it is just spotting based. If a unit is revealed to the unit then they fire at it. So I think the entire issue with vehicles not firing is purely a spotting issue and I will go about enhancing how CF_BAI/detect works so that vehicles have long-range optical vision and such and do a bit of rebalancing for alpha2.
  11. One of the changes we made to GAIA in MCC in our branch of the mod changes the algorithm for vehicles so that they actually set a waypoint some 300-700m away from the target instead of just charging it. Somewhere along the line in Arma 3's patches, BI changed the behavior so they pushed on to the waypoint and it made all vehicles bull rush people. By shortening the generated waypoint we improved the charging in. The problem is they still don't actually shoot and they certainly don't face themselves well to minimize damage. Something like this algorithm is probably part of the solution.
  12. Few things on how to address them: I think ACE would be a decent place to change the damage model for the tanks to increase their survivability, they already have a framework for adjusting damage it just needs some tweaks for tanks and vehicles. Some of them are too tough at the moment as well. The accuracy of tanks is something I think I can deal within CF_BAI. I can take a unit and run through our subskill pipeline and if they get in a tank drop their accuracy down. This is basically what CF_BAI does, we fix the balance of the game being all wrong with the unit's skills. Whether this works or not is another matter because the big issue with explosives is that a miss might still be close enough to splash hit. There has been a consistent issue with RPGs and other such explosives used on infantry where it misses by 1m and hence just ends up hitting. Not sure how much more control we can assert on that one even with adjustments in the config for reducing accuracy but we can certainly impact it to an extent. CF_BAI/detect probably doesn't currently extend the spotting range of tanks properly but it is on the todo list. I think we can solve that one, it is just a matter of fixing the raycast to work with vehicles, the distance of vision extended and facing narrowed for vehicles. JSRS maybe could fix with some increased range of volume for tanks? CF_Woodland could potentially make a lot more branch cracking noise for vehicles moving in woodland, sounds like a good feature for it to have. The AI runs into trees a lot and then they retry the same motion over and over. VCOM AI has been working on this for a while but I think it is mostly focussed on dealing with road transit. Do we have sufficient low level control of a vehicle to choose locations to move to? If so I think my new CF_BAI/terraincache might be able to help produce a performant pathfinding algorithm, I have considered using it for better cover finding for infantry and maybe we can use a preprocessing of the map for vehicle motion. Lot of pieces to test and work out on this one but it depends on just how much control we have and it is a big job. Enemies engaging from vehicles is a problem in general. I thought much of it was spotting due to the big reduction in distances BI did with tac ops but having fixed that with CF_BAI/detect it is now just a problem with them firing. There is a great video done by ShockNAwe that shows just how messed up it is with two opposing enemies just 30m from each other just sat there. I think it could, in theory, be scripted around by using the knowsAbout and assigning targets or perhaps a suppressive fire or doFire commands. Something in there might work to get them shooting. I am definitely interested in anyone who works on this as I want a solution, but I can also fix this and potentially make it a module in CF_BAI but we don't have anything like this currently. The AI don't seem to engage vehicles that don't have people in them. Since a lot of the explosive stuff is never fired at troops on foot it is a good way to avoid getting "RPGed". This was purposefully changed by BI a while ago and not a change I agree with. I can't think of a nicely performant way to fix it beyond using nearestObjects targeting the other sides vehicles and shoot much like the solution to 7 would likely work. These two definitely go together as features and I can't imagine this is too difficult. I think CF_BAI with a few changes addresses some of it, it is open source so if you want to come give me a hand with those bits then more than welcoming of pull requests https://github.com/BrightCandle/cf_bai
  13. Everything is a visionTest, where a visiontest is 2 numbers, the first is the probability (0-1) that the unit will see and the second is how much knowledge they would gain if they did see it (also 0-1 mapped to 0-4 for reveal/knowsAbout purposes). For the environment, the tests are dayNight, rain and fog. For the player, it is stance, movement and foilage For the AI, it is awareness, binoculars, optics, nightvision and spotTime For an AI looking to a player it is existingKnowledge, facing, distance and a ray cast
  14. Not yet, but it will be in the future. Only the player side is currently done and I need settings for the AI, environmental and player to AI tests.
  15. A major update today. I have added 2 new modules to CF_BAI. The first is detect which runs a series of vision tests taking into account the environmental conditions, the AIs situation and the players situation and combines it all into a chance to spot and gain an amount of knowledge. AI can potentially spot players over 900m away without optic assistance and potentially many KM's away with optics. This is a workaround for the 1.84 especially but also every release of the game since Apex that has reduced the AIs maximum unaware spotting distance to a very unrealistic 320m on maximum spotting skill. AI will look at you when they gain knowledge but it is otherwise the only way to see it working is via the units knowsAbout. The second module is suppression. The suppression module takes the existing suppression value of the unit and then converts this into an impact on the units subskills so their shooting and spotting are all reduced by configurable amounts. In addition, the default suppression value decays within a second once shots have stopped and this is far too short so the system has a hold and decay that can be up to about a minute before the unit is fully recovered from being suppressed. This isn't quite as widely detecting as TPWCAS but it also has nice low CPU impact and contains no behavior just skill reductions allowing ASR_AI and VCOM AI to do what they do in response to contact. Both modules are still a bit experimental as we haven't run them for more than a few weeks and there are some flaws with the detect module that are already known that need do addressing (details on the website and on the Steam workshop). Normally we would test and develop for much longer but I think what we have is already very usable and important to game balance especially with 1.88 breaking TPWCAS, so I decided to push out a release that we believe is bug-free but far from perfect in its implementation right now. For other mod makers that might want to do subskill dynamic adjustments, we have also included an API consisting of 4 functions that allow setting subskills without impacting on other mods that also adjust it or for setting the units base skill. I am not so sure the way this will be utilized is via CF_BAI but we now have an approach that CF_BAI uses internally which is exposed for feedback. The best place for it is likely CBA if other mod makers want to use it.
×