Author Topic: Autonomous Combat League - Bot Workshop  (Read 1360 times)

cupid_the_conqueror

  • Global Moderator
  • Experienced Member
  • *****
  • Posts: 89
  • Likes: +0/-0
  • A filthy casual
    • View Profile
    • the truth of rawbots
Re: Autonomous Combat League - Bot Workshop
« Reply #30 on: March 25, 2016, 03:31:59 pm »
what in the seven hells?! I've never seen a nan output in rawbots before!? maybe try two different parts? I dunno, thats bloody strange man.
At the end of the day, love conquerors all.

MarvinMan

  • Automation Expert
  • Experienced Member
  • ***
  • Posts: 236
  • Likes: +4/-0
    • View Profile
Re: Autonomous Combat League - Bot Workshop
« Reply #31 on: March 25, 2016, 03:49:02 pm »
NaNs usually show up as soon as you load some code that has a loop with no external input or pre-defined starting condition. The classic case was using an integrator and sin/cos/tan function to calculate the inverse back in the days before XFM added those functions. Amusingly, the last time I had a problem with NaNs was when I was building the turret for my first combat bot.

This time things are weird as the bot will run happily for a while before the NaNs appear, and the MIN function isn't filtering them out. Normally, the VP hexes can handle NaN values properly and you only have a problem when trying to drive an actuator with NaN.

Here's the blueprint, if anyone wants to see if they can replicate the problem:
https://www.dropbox.com/s/lp37fg5eqpszcpk/bp_NaNomatic?dl=0

MarvinMan

  • Automation Expert
  • Experienced Member
  • ***
  • Posts: 236
  • Likes: +4/-0
    • View Profile
Re: Autonomous Combat League - Bot Workshop
« Reply #32 on: March 25, 2016, 09:13:30 pm »
Well that's odd. It appears that MIN/MAX functions only handle NaNs properly if the NaN appears on the in_a input.

Here's a screenshot of the new target detection system in action that I think looks pretty cool.

PressureLine

  • XFM Developer
  • Administrator
  • Experienced Member
  • *****
  • Posts: 192
  • Likes: +4/-0
    • View Profile
    • Rawbots Club Forums
Re: Autonomous Combat League - Bot Workshop
« Reply #33 on: March 25, 2016, 11:26:59 pm »
I've only noticed NaNs coming from ACOS/ASIN/ATAN functions, limiting their input to 1 helps a lot with that.
"Never assume that anything you've never seen before is benign. Most particularly not something that says 'MARINES' on it."

PressureLine

  • XFM Developer
  • Administrator
  • Experienced Member
  • *****
  • Posts: 192
  • Likes: +4/-0
    • View Profile
    • Rawbots Club Forums
Re: Autonomous Combat League - Bot Workshop
« Reply #34 on: March 26, 2016, 05:07:01 am »


Combat systems functional, now for lift and terrain avoidance :)
"Never assume that anything you've never seen before is benign. Most particularly not something that says 'MARINES' on it."

MarvinMan

  • Automation Expert
  • Experienced Member
  • ***
  • Posts: 236
  • Likes: +4/-0
    • View Profile
Re: Autonomous Combat League - Bot Workshop
« Reply #35 on: March 26, 2016, 11:53:57 am »
Is that another lightweight class entry? It'll be interesting to see how stable it is after taking a bit of damage.

PressureLine

  • XFM Developer
  • Administrator
  • Experienced Member
  • *****
  • Posts: 192
  • Likes: +4/-0
    • View Profile
    • Rawbots Club Forums
Re: Autonomous Combat League - Bot Workshop
« Reply #36 on: March 28, 2016, 05:22:37 am »
Quick crosspost:

Ok, minor rules change to spread the weight classes around a bit:
  • Lightweight: <=25 parts
  • Middleweight: 26-50 parts
  • Heavyweight: 51-75 parts
  • Unlimited: >75 parts
There is a bit of reasoning behind this, so bear with me: It makes each weightclass a bit broader, and better represents (imo) the way combat power escalates with increased partcount. It should also have the effect of making any potential swarmbots a bit more competiticve, since now a combo of 2 max weight light bots counts as a single middleweight entry, and not a heavyweight.
"Never assume that anything you've never seen before is benign. Most particularly not something that says 'MARINES' on it."

PressureLine

  • XFM Developer
  • Administrator
  • Experienced Member
  • *****
  • Posts: 192
  • Likes: +4/-0
    • View Profile
    • Rawbots Club Forums
Re: Autonomous Combat League - Bot Workshop
« Reply #37 on: April 16, 2016, 11:52:43 pm »
Well that's odd. It appears that MIN/MAX functions only handle NaNs properly if the NaN appears on the in_a input.

Here's a screenshot of the new target detection system in action that I think looks pretty cool.


I had a play with the NaNomatic blueprint, and it is a pretty cool idea, just needs a better design for the turrets, capable of 90 traverse and elevation control and you've got a solid little bot there.
"Never assume that anything you've never seen before is benign. Most particularly not something that says 'MARINES' on it."

MarvinMan

  • Automation Expert
  • Experienced Member
  • ***
  • Posts: 236
  • Likes: +4/-0
    • View Profile
Re: Autonomous Combat League - Bot Workshop
« Reply #38 on: April 18, 2016, 10:39:11 am »
That bot is actually the development platform for the RedEye, which is still running nearly identical code. The RedEye was a complete re-build, but the only real layout change was moving the cannons from the back to the front; everything else was just a case of making the systems lower and more compact.

PressureLine

  • XFM Developer
  • Administrator
  • Experienced Member
  • *****
  • Posts: 192
  • Likes: +4/-0
    • View Profile
    • Rawbots Club Forums
Re: Autonomous Combat League - Bot Workshop
« Reply #39 on: April 18, 2016, 11:28:42 pm »
Got about halfway through coding a dual turret bot last night. The code is getting pretty monstrous already, and the turrets can't even aim yet :-\
"Never assume that anything you've never seen before is benign. Most particularly not something that says 'MARINES' on it."

MarvinMan

  • Automation Expert
  • Experienced Member
  • ***
  • Posts: 236
  • Likes: +4/-0
    • View Profile
Re: Autonomous Combat League - Bot Workshop
« Reply #40 on: April 19, 2016, 10:29:41 am »
I would have thought multiple turrets would be fairly straightforward to do. I'd probably just duplicate the turret/target acquisition system, then have the navigation system drive towards whichever target is closest. I suppose the tricky part could be making sure that turrets fire at different targets.

That could be as simple as having each turret only cover a 180 degree range, but you could try to prevent the turret locking onto targets within a given distance of what the other turret is firing at.

PressureLine

  • XFM Developer
  • Administrator
  • Experienced Member
  • *****
  • Posts: 192
  • Likes: +4/-0
    • View Profile
    • Rawbots Club Forums
Re: Autonomous Combat League - Bot Workshop
« Reply #41 on: April 19, 2016, 10:57:37 am »
I didn't feel the need to arbitrarily limit the turrets from aiming at athe same target, they simply target something in the view of one of the cameras that is assigned to them that can see the most parts. The chassis 'aims' itself at the target selected by the camera that can see the most parts (it has access to all of the camera outputs)



Nearly there :D
"Never assume that anything you've never seen before is benign. Most particularly not something that says 'MARINES' on it."

MarvinMan

  • Automation Expert
  • Experienced Member
  • ***
  • Posts: 236
  • Likes: +4/-0
    • View Profile
Re: Autonomous Combat League - Bot Workshop
« Reply #42 on: April 19, 2016, 11:09:46 am »
I suppose most of the time it's better to be concentrating fire on a single target. The advantage of having them lock onto different targets would be to prevent another bot sneaking up on it. Multiple independent turrets could also be useful against swarms or could help prevent the situation where the bot keeps firing at a heap of parts.

It could be interesting to have a bot with a big main gun aimed at the primary target, then a small secondary turret to attack bots approaching from a distance.

PressureLine

  • XFM Developer
  • Administrator
  • Experienced Member
  • *****
  • Posts: 192
  • Likes: +4/-0
    • View Profile
    • Rawbots Club Forums
Re: Autonomous Combat League - Bot Workshop
« Reply #43 on: April 19, 2016, 11:15:51 am »
Yeah. This case deals more with the fact that the turrets share a part of their arc, so if they both can shoot at the same target they probably should be, but where the target is far enough off to one side (for whatever reason) that one of the turrets cannot shoot at it there may be a secondary target that is worth shooting at.
"Never assume that anything you've never seen before is benign. Most particularly not something that says 'MARINES' on it."

PressureLine

  • XFM Developer
  • Administrator
  • Experienced Member
  • *****
  • Posts: 192
  • Likes: +4/-0
    • View Profile
    • Rawbots Club Forums
Re: Autonomous Combat League - Bot Workshop
« Reply #44 on: April 19, 2016, 01:45:44 pm »
And done :)

"Never assume that anything you've never seen before is benign. Most particularly not something that says 'MARINES' on it."