So a tendency for the AI is to use its commander offensively, and with no backup or sense of self preservation. It seems to be fairly common for it to walk alone directly into the line of fire and die. This isn't very exciting. Just now I had a game where I decided to play two AI off of each other, they spawned pretty close to each other and within a few minutes the game ended like this, in favour of red only because red had a few extra tanks hanging around the base. And I know it's been a thing since at least supcom1, in fact I'm pretty sure it was how I won my very first game of supcom back when the demo was fresh. It's rather ludicrous that it's still a thing. I know the game isn't finished but I want some soothing words of reassurance that it's being looked at.
The AI has no commander self preservation yet. I may add some soon, but I really wanted to wait until I had neural network code in so the commander can have one.
It's a silly feature. It gets fixated on building metal and runs through my base to get to an unclaimed mex spot. Then it builds metal. In the middle of my base. With everything shooting at it. As if I'm the problem. On the other hand, the AI built it's very first proxy today! I'm so proud. Growing up too fast. It was even doing a turret creep in my proxy when it built it. Happy days.
I can understand not wanting to fiddle too much but really feels like some sort of temporary confinement for the commander to close proximity of spawn is necessary, it's pretty difficult to test the AI when it often sends itself on a suicide mission early on in the game.
I have noticed this, playing the AI on planets size 3 or greater seems to fix this issue for the most part. AI will straight up roll in and murder you comm rush style, if you build anything near his base early game.
This is the second most annoying bug when playing a singleplayer skirmish game in the 62165 build. Can you give us any ETA on that neural network for the AI? Thank you very much.
I have a fix in place that should stop this. It wont stop the AI wanting to uber cannon units attacking its base, but it should stop them going walkabout in your base.
Unfortunately, commander suicide behaviour still exist and currently is the most annoying thing in PvE game. Yesterday we've played 5 games in a row with different AI difficulty settings from Hard to Absurd in a setup like: 2 players vs 3-4 AIs. In ALL of the games there was at least one AI commander that ended its existance by walking right into the players defence grid and attacking it (i had many turrets and vehicles+air units in that place). It was walking with its other battle units as i could remember and there was NO visible conditions for such action because its main base were untouched (the game was in build/defence phase for us). One thing to note: i've seen the situation once where commander appeared near defence grid, destroyed one tower and walked away with 10%hp left. If tower won't go down quickly commander possibly could be killed like others. Seems like the bug to me, cause in most cases the AI commanders are somewhat protected. Hope devs will read that and try to fix, cause the PvE is suffer alot from this bug.
Changes in the PTE (the public testing environment) might prevent this: AI: - AI now expands faster - AI can keep track of last time it had vision on a unit and use this info when deciding what to do - Fix for duplicate AI build item - AI will be less likely to build on top of metal In the past the AI might have scouted the defence grid and forgot it was there before sending his commander. Now it should remember it and thus not send his commander to his death. Not sure if it fixes this problem but I think it might
