Unit Adjacency Responses

Discussion in 'Backers Lounge (Read-only)' started by ViolentMind, September 14, 2013.

  1. ViolentMind

    ViolentMind Active Member

    Messages:
    394
    Likes Received:
    186
    *It's possible that this has been discussed before, however I have searched a bit and couldn't find anything.

    One thing that I've always found annoying about RTS games was the fact that, as in real life, units do not respond to what is happening to the units right next to them in any intelligent way. The prime example of this is when there are a group of units sitting idle (maybe at a production gathering point, or because their order queue has been completed), and an approaching enemy force starts picking off those units one by one until they are all dead. This happens because without direct orders to respond to the attack, the units will only individually engage enemies that come into their field of vision. The result is, instead of fighting as a group against the enemy mass, they fight the enemy group one at a time and get slaughtered. In any real battle, when a threat is discovered by one person, the ones around them respond accordingly. You see this even in ants, when danger is discovered, it is communicated to those in the vicinity and they all respond to the threat as if it is happening to them directly. I've seen discussions about Area Commands for patrol groups (among other things), which I think is a great feature! However, what about an area of alert/awareness for auto-responding to an enemy threat? This area of response could simply equal each respective units' area of vision. Each unit that could "see" a unit being attacked would immediately respond to the attacking group as if they were being attacked themselves. In this way, a group of assembled units would do the amount of defensive damage corresponding to their numbers a bit more accurately, and the player would not return to find his once powerful idle group of units completely decimated by a much smaller force, just because he/she didn't have the bandwidth to actively issue them a group order. The group may still be destroyed even with this unit adjacency awareness/response feature, but at least it wouldn't be such an unrealistic one sided victory with a disproportionate amount of unit loss. A similar response for idle grounded air units would be to become airborne again to avoid being destroyed by ground forces. In my opinion, this becomes more important in larger scale games when the chances increase for unit groups to sit idle.
  2. Ringworm

    Ringworm Active Member

    Messages:
    245
    Likes Received:
    81
    Interesting Idea.

    Something (else...) for Sorian to consider?
  3. eroticburrito

    eroticburrito Post Master General

    Messages:
    1,633
    Likes Received:
    1,836
    I'm pretty sure that units respond if an enemy attacks something close by and the enemy is in their direct line of sight/range?

    It would be better if a unit saw a nearby unit/structure under attack and automatically went to defend if said structure/unit was in their line of sight.
    I'm not sure what the plans are for unit formations but there are two systems which spring to mind;
    - Starcraft - 'Mobs' of units rushing around; the current system. Leads to heavy micro, a lot of kiting - of which I am not a fan. But being so close to one another allows them to respond easily to threats - as they all see the same thing at once.
    - SupCom - Formations of units which do as you say and start getting picked off around the edges unless you intervene.

    What's needed is to take the best from both systems - for units to respond to a threat but also not run around like an angry mob, but in formation like an army of killing machines.
    To do this they need to communicate as you suggest (like ants), so that a nearby unit can 'call' other units within their LOS to assist.

    What might be useful to avoid kiting with units that respond to a threat is a player-determined "Guard Zone", with a circle that you could control the radius of in order to prevent your units being lured into your opponent's point defences.
    However this too would need to be controlled so that units didn't just stop at the edge of the zone and get hammered by artillery, but returned to the middle of the zone/resumed patrols if they can't fire upon an enemy.

    Edit: Also agree about air needing to take off if they're getting shelled/see an enemy coming. One of the most retarded things about SupCom.
    Last edited: September 14, 2013
    carpetmat likes this.
  4. carpetmat

    carpetmat Member

    Messages:
    66
    Likes Received:
    23
    I agree, but i wouldn't expect to want my army moving itself in range all time. Other players could bait more of your units into range of artillery. What if in addition to the guard mode burrito explained, only units grouped via "ctrl+number" responds as a group and not an individual.

    Example:
    A stationary group of tanks given the guard order. Once an enemy unit is within visual range of at least one tank, all tanks begin to move into firing range, however they don't pursue enemy tanks that leave that initial visual range.
  5. exterminans

    exterminans Post Master General

    Messages:
    1,881
    Likes Received:
    986
    I recommend all four of you take a visit at Sorians blog and take a look at the neural network based unit AI he already wrote for SupCom 2 (was only active for AI players in SupCom 2).

    The behavior you are asking for, is one of the implicit behaviors the unit AI would show. This even includes such seemingly counterintuitive moves such as advancing with reckon units in such a scenario to reveal additional targets, as well as automatic retreat if the losses would have been unbearable.

    Explaining WHY the units would do this is a bit difficult though, hence read the blog posts on that topic if you want to know how it works.

    A simplified summary:
    The unit AI gathers statistical informations about it's surroundings. It matches these informations against an algorithm which was calibrated with past outcomes and the algorithm tells the unit the optimal behavior in that situation for the current unit type. The unit is not aware of other units in the area as individual entities, but rather about their potential as an abstract number.
    The larger the area the statistics are gathered on, the further the unit CAN travel to assist, but it will only do so if it is "sure" that it will help.
    Last edited: September 14, 2013
    eroticburrito and carpetmat like this.
  6. carpetmat

    carpetmat Member

    Messages:
    66
    Likes Received:
    23
    Thanks for the heads up, looking for it now!
    Found it, i need to read this blog more, interesting topics. Thanks for pointing this out exterminans!

    I think that level of response might make things too automated for the player. It would be super helpful, but probably too much for competitive play. I would imagine it to be a feature of a "guard" order. There by requiring player attention to initiate.
    Last edited: September 14, 2013
  7. ViolentMind

    ViolentMind Active Member

    Messages:
    394
    Likes Received:
    186
    Yeah, my thought was that the units responding to "guard" would only really move to attack the enemy unit(s) which initiated the attack on their group. They would only move to the point where they could shoot back. They either would not give chase if the enemy unit retreated, or they would only pursue as far as the area command border was defined (if it was defined). I could see this behavior only applying to grouped units as well. I don't think that it needs a complex algorithm to define how far to chase, or if it should chase the enemy. Once the enemy retreated out of the defined area in the area command, or if no area command was defined, the units in the group would simply return to their original positions. That way, you would not have an excessive amount of AI intervention to fight your battles, the enemy could not just walk in and pick off all of your edge units until your whole group was gone, and all of the units that you worked so hard to create would defend to the best of their ability until you could once again take control.

    I think that this approach would create less micro from both the attacker and defender. This obviously assumes that the attacking and defending units had the same range of visibility, if not, and the attacker had longer ranged units and kept them at range, then he deserves to slaughter the idle group without losses. The attacker, however, knowing how the like ranged units would probably respond, wouldn't be tempted to kite around as much trying to take advantage of no unit adjacency awareness (which is in my opinion just a tactic designed to take advantage of poorly implemented unit behavior). The defender can also rest a little easier knowing that his units will respond, maybe not as well as with human control, but at least they would not be snuffed out without dealing some relevant damage.

    Of course, an easier alternative to all of this fancy unit behavior would be to just set your group of units to default to a "ball" formation. That way, at least multiple units would probably engage the enemy at the same time, no matter which direction the group was approached from.
    Last edited: September 16, 2013
  8. exterminans

    exterminans Post Master General

    Messages:
    1,881
    Likes Received:
    986
    Thats the whole point. Micro as in direct unit control isn't to be the focus of the game at all. That's what you have games like Starcraft for where APM matters above all, and everything, including a proper scale of the map, is sacrificed for that "feature".

    PA and it's predecessors all have been about smart decisions on the tactical level and above, less so with TA since it contained a lot of exploitable micro, but even more so in SupCom to not to divert from the scale of the game.

    You give mostly high-level orders, and your units execute them in the most efficient manner. They know exactly which target to shoot and when to start "dancing". Gaining an advantage is not so much about micromanaging the individual units, but choosing attack vectors, movement paths, unit composition and alike before and seeing how it turns out while you are able to attend at a different site already.

    It's still quite time consuming if you "only" require to check back on your units every 10 seconds or so to issue new target locations so the enemy won't be able to outsmart you with real flanking maneuvers. (Which is NOT part of the unit AI, that's something you have to do all by yourself.)
    CommieKazie likes this.
  9. ViolentMind

    ViolentMind Active Member

    Messages:
    394
    Likes Received:
    186
    So, I'm confused...are you arguing for better automated defensive response behaviors for the type of idle unit group situations that I described, or against them? : )

    I completely agree with you about micro not being the focus of the game, and overall strategy and tactics being the focus! That's essentially my point in starting this discussion. That's what I want and partly why I backed this game. Having said that, the scale of this game will blow away SupCom with the larger multi-planet setups. All the more reason to build in smarter unit behaviors. I don't mean to suggest that the AI should take control of your units and start making offensive decisions for you about where to send them. I am just hoping that they become more reliable in defending themselves when you are not around to babysit them, and that players won't be able to take advantage of "dumb" units that sit around and wait to be slaughtered. In a similar fashion, I think that even your fixed defenses (like turrets) should respond more intelligently when a nearby friendly unit, or building, is attacked. If the attack comes to the friendly object within their field of vision, then they should turn to face the direction of the attacking units, instead of waiting until they get shot to start turning their guns to fire back (as it currently works in the Alpha). After all, not being able to micro all of your units to defend properly doesn't represent poor strategy, it's just a symptom of poor APM. I'm hoping that Jon and the gang don't penalize us for that, like Blizzard has chosen to do. PA Uber Alles!!!

    PS - Uber needs to name the most powerful unit, or offensive structure, in the game Uber Alles! haha

Share This Page