Difference between revisions of "User:Skilgannon"

From Robowiki
Jump to navigation Jump to search
(bring to present)
(Add Kd-tree)
 
(7 intermediate revisions by 3 users not shown)
Line 1: Line 1:
I'm a student, studying Mechanical Engineering at University of Cape Town, South Africa. I currently have a range of (mostly competitive) bots:
+
I'm a student of life, having lived in the US, South Africa, Switzerland and now Germany. Robocode was the first thing to get me interested in software development at the age of 16, and now 10+ years later I do similar work (computer vision and machine learning) professionally, although in C++ not Java. Somewhere along the way I got a B.Sc.Eng in Electro-mechanical Engineering, and a M.Sc.Eng in Robotics, both from the University of Cape Town.
  
 +
I currently have a range of competitive bots:
 
1v1:
 
1v1:
  
 
* Nano:
 
* Nano:
** [[Machete]] - Uses [[Multimode]] - [[Stop And Go]] which switches to [[Random Movement]] - and nanonized [[Wall Smoothing]], with a [[Random Lead]] gun. Not as strong as I hoped.
+
** [[Yatagan]], which I wrote together with [[User:Sheldor]], the top nano which does actual learning (ie. no offline generated tables which are matched by names to enemies)
** [[Olek]] - Uses the [[Musashi Trick]] coupled with a nano [[Pattern Matcher]]. No [[Wall Smoothing]] (not enough [[Code Size]]. Also not as strong as I had hoped.
 
  
 
* Micro:
 
* Micro:
 
** [[Toorkild]] - A [[Multiple Choice]] [[Pattern Matcher]] inspired by the competition, [[Kev]]'s [[Needle]]. Number 1 in Micro.
 
** [[Toorkild]] - A [[Multiple Choice]] [[Pattern Matcher]] inspired by the competition, [[Kev]]'s [[Needle]]. Number 1 in Micro.
** [[Waylander]] - A  [[Pattern Matcher]] - currently No. 2 or 3 in the [[Micro Rumble]]. It was my first attempt at a competitive bot.
+
** [[Waylander]] - A  [[Pattern Matcher]] - once no. 1 in the [[Micro Rumble]]. It was my first attempt at a competitive bot. - (RETIRED)
** [[Connavar]] - A [[Guess Factor]] bot, using [[Waylander]]'s movement, and a gun based (slightly) off of [[Falcon]].
+
** [[Connavar]] - A [[Guess Factor]] bot, using [[Waylander]]'s movement, and a gun based (slightly) off of [[Falcon]]. - (RETIRED)
  
 
* Mini:
 
* Mini:
** [[Cunobelin]] - A [[Wavesurfer]] based loosely off of [[BasicSurfer]]
+
** [[CunobelinDC]] - The gun from Toorkild, but with a [[Dynamic Clustering]] surfing stats algorithm. Number 1 in Mini.
  
 
* Mega:
 
* Mega:
** [[DrussGT]] - A [[Wave Surfer]] using [[Visit Count Stats]] for the movement, and [[Dynamic Clustering]] for the gun. Currently No. 1 overall.
+
** [[DrussGT]] - A [[Wave Surfer]] using [[Visit Count Stats]] for the movement, and [[Dynamic Clustering]] for the gun. Currently No. 1 in the 1v1 unrestricted section.
** [[Stormrider]] (retired) - [[Wave Surfer]] based off of [[DrussGT]], but using [[Dynamic Clustering]] for the movement. It shares [[DrussGT]]'s gun. While in the rumble it was No. 5 overall.
+
** [[Stormrider]] ( [[Wave Surfer]] based off of [[DrussGT]], but using [[Dynamic Clustering]] for the movement. It shares [[DrussGT]]'s gun. While in the rumble it was No. 5 overall.  - (RETIRED)
 +
** [[EnergyDome]] - A bullet shielding bot with precise prediction, and a bunch of virtual enemy guns to handle common configs. One day I would love to add a learning gun, just to see if it would work. I think there is actually a bunch of potential still left in the bullet shielding idea.
  
 
Melee:
 
Melee:
 
+
* Mega:
 +
** [[Neuromancer]] - So far, the only bot to attempt to simultaneously surf all melee waves, including the ones not fired at it. Uses a [[Dynamic Clustering]] [[Play It Forward]] gun. Super strong in Melee (#1 for a few years), and slowly closing the gap with DrussGT in 1v1. Still a bunch of missing features that I'd love to add, both for melee and 1v1. I'd also like to make it work for teams!
 
* Mini:
 
* Mini:
 
** [[Snaga]] - A currently unreleased bot using a segmented [[Play It Forward]] gun I built originally for [[Twin Duel]], using [[Minimum Risk Movement]]
 
** [[Snaga]] - A currently unreleased bot using a segmented [[Play It Forward]] gun I built originally for [[Twin Duel]], using [[Minimum Risk Movement]]
Line 27: Line 29:
 
* [[Twin Duel]]:
 
* [[Twin Duel]]:
 
** [[NightAndDay]] - Two bots, Night and Day, working together. They use a segmented GF gun with a careful choice of attributes to enable hitting non-perpendicular movement. They use [[Minimum Risk Movement]]. The only difference between them is the color.
 
** [[NightAndDay]] - Two bots, Night and Day, working together. They use a segmented GF gun with a careful choice of attributes to enable hitting non-perpendicular movement. They use [[Minimum Risk Movement]]. The only difference between them is the color.
 +
 +
Utilities:
 +
* A really high performance [[Kd-Tree]] that is used in a number of bots including DrussGT and Neuromancer, check it out [[User:Skilgannon/KDTree|here]]
 +
 +
[[Category:Bot Authors|Skilgannon]]

Latest revision as of 21:58, 20 September 2017

I'm a student of life, having lived in the US, South Africa, Switzerland and now Germany. Robocode was the first thing to get me interested in software development at the age of 16, and now 10+ years later I do similar work (computer vision and machine learning) professionally, although in C++ not Java. Somewhere along the way I got a B.Sc.Eng in Electro-mechanical Engineering, and a M.Sc.Eng in Robotics, both from the University of Cape Town.

I currently have a range of competitive bots: 1v1:

  • Nano:
    • Yatagan, which I wrote together with User:Sheldor, the top nano which does actual learning (ie. no offline generated tables which are matched by names to enemies)
  • Mega:
    • DrussGT - A Wave Surfer using Visit Count Stats for the movement, and Dynamic Clustering for the gun. Currently No. 1 in the 1v1 unrestricted section.
    • Stormrider ( Wave Surfer based off of DrussGT, but using Dynamic Clustering for the movement. It shares DrussGT's gun. While in the rumble it was No. 5 overall. - (RETIRED)
    • EnergyDome - A bullet shielding bot with precise prediction, and a bunch of virtual enemy guns to handle common configs. One day I would love to add a learning gun, just to see if it would work. I think there is actually a bunch of potential still left in the bullet shielding idea.

Melee:

  • Mega:
    • Neuromancer - So far, the only bot to attempt to simultaneously surf all melee waves, including the ones not fired at it. Uses a Dynamic Clustering Play It Forward gun. Super strong in Melee (#1 for a few years), and slowly closing the gap with DrussGT in 1v1. Still a bunch of missing features that I'd love to add, both for melee and 1v1. I'd also like to make it work for teams!
  • Mini:

Teams:

  • Twin Duel:
    • NightAndDay - Two bots, Night and Day, working together. They use a segmented GF gun with a careful choice of attributes to enable hitting non-perpendicular movement. They use Minimum Risk Movement. The only difference between them is the color.

Utilities:

  • A really high performance Kd-Tree that is used in a number of bots including DrussGT and Neuromancer, check it out here