Difference between revisions of "User:Dsekercioglu"

From Robowiki
Jump to navigation Jump to search
(Update user page)
 
(14 intermediate revisions by the same user not shown)
Line 1: Line 1:
 
==Me==
 
==Me==
My name is Doruk Sekercioglu. I am a 13 years old Turkish student in high-school. Doing robocode as a hobby. Probably the youngest one here(There are bots older than me=)). I started robocode last year and I thought that head-on targeting was the best targeting method and circling around the enemy was the best movement because head-on targeting couldn't hit it =). I am checking the wiki like 6 times a week and improving my bots.
+
:I am a Turkish student currently in high school. My teacher introduced me to Robocode and it probably is the only reason I do programming today.
 +
:Over the time, I've come to realize what a complicated optimization problem robocode was.
  
==My Bots==
+
:Since the July of 2020, I have spent more time learning other languages, most of them being lower level and I realized how inefficient most of my code was. Although I don't really feel like using Java anymore, I don't think I can possibly give up on robocode just yet. At least not before I give a final shot with Neural Networks ;)
==Extends Robot==
 
;TheFieryBot: Extends Robot, Spin Gun, Square movement
 
;FireRate: (My plan was to make this bot beat every bot but I thought that the only bots were the sample bots) Extends Robot, Smarter Ramming
 
;ParallelFire: I saw Diamond vs Shadow on Youtube and I said "Wow I should make a robot like this" Result: Fire, Go back, Go ahead, Fire again
 
There are like 25 other bot I will skip them.
 
  
==Advanced Robots==
+
==Bots==
;Golem, Cyclops, Phoenix, DeadKnight: All of them keeps the track of enemy energy and moves to dodge
+
:I have made many robots, mostly experimental and most of them are retired.
;M103, T62A, Obj742?(I don't remember): All of them circles the enemy and fires at the enemy.
 
;Note: I thought that nobody could hit any of them(I thought that Diamond and Shadow was firing randomly =))
 
==After Downloading Rapture==
 
Rapture is not a very strong bot (601 currently) but it had pattern matching and just beat my bots in a second.
 
Then everything changed about robocode.
 
;TheBeast: Some sort of random movement, Tried to do linear targeting. UNRELEASED
 
;Tomahawk: My first bot in the rumble bot but I needed to wait for 5 months to put it into roborumble because I didn't know how to.
 
;(Thanks very much to [[Beaming]] and [[Skilgannon]] they helped me a lot)
 
Tomahawk used DC-PIF for everything but I didn't normalised the weights so while distance was weighted 1200; velocity was 16.
 
==Things I Invented When I Was Improving Tomahawk==
 
;Random Movement
 
;MinimumRisk Movement
 
;Angular Minimum Risk Movement(Instead of looking distances it looked at angular difference).
 
;Linear Gun(It was changing it's fire power to shoot)
 
;Circular Gun(Just a complicated formula and needed to change fire power to hit the enemy)
 
;Mean Linear/Circular
 
;[[Laser Targeting]]
 
;[[Virtual Guns]]
 
;GF -1/0/1 gun(Precise MEA)
 
;Non-Segmented VCS Gun
 
;KNN with PIF, GF and Displacement Vectors.
 
;KNN with kernel density but I didn't know what kernel density was.
 
;Wave Surfing by looking to the graphics of [[Diamond]]
 
  
==After Finding Robowiki==
 
I saw everything that I found and a lot more was here.
 
I just started to read it everyday when I get bored.
 
  
==Good Bots==
+
:[[ColdBreath]]:
;[[WhiteFang]]: My first good bot. I wrote the whole code on my own but I used [[Rednaxela]]'s 3rd gen Kd-tree.
+
:My first competitive NN bot. Uses [[Roboneural]]. Reached a ranking of 55 with only 1 predictor for both movement and gun.
Best Ranking: 114
 
;[[Husky]]: Simple bot set on [[BasicGFSurfer]] with a DC-GF Gun and Segmented VCS movement.
 
Best Ranking: 86
 
;[[Havoc]]: Neural bot set on [[BasicGFSurfer]] with a MLP Gun and a MLP Movement. RETIRED(Wrong Backpropogation)
 
Best Ranking: 101 (When using KNN)
 
;[[Zoom]]: NN  bot set on [[Havoc]] with a NN special to robocode. My current bot. RETIRED(Wrong Backpropogation)
 
Best Ranking: 120
 
;[[Wyvern]]: VCS bot which I made from modular [[BasicGFSurfer]]. Creates random [[VCS]] predictors.
 
Best Ranking: 67(Because of a performance enhancing coding mistake)
 
;[[ColdBreath]]: Pure NN Bot. Very messy code built on Modularised [[BasicGFSurfer]].
 
Best Ranking: 56(It is falling down because I test my bots against [[ColdBreath]])
 
;[[Oculus]]: Pure NN Bot. Uses multiple NN's in movement. Very clean code and modular design. Very easy to add new things.
 
Best Ranking: 56(Once it passed [[ColdBreath]])
 
  
==Projects==
+
:[[Oculus]]:
:[[Roboneural]]: There are Neural Networks designed for robocode in this project and [[SwordOfFire]], [[ColdBreath]] uses it.
+
:[[Oculus]] reached a ranking of 35 but was later removed temporarily since it was too slow.  
:[[Grapher2D]]: Using an Auto Encoder, this bot will try to show the 5D movement profile in 2D.
 
:[[OculusRammer]]: Couldn't find a good name yet but I want this bot to be the best rammer.
 
  
:[[Category:Bot Authors|Dsekercioglu]]
+
:[[WhiteFang]]:
 +
:[[WhiteFang]] is my third robot to be released. It's now in top 10 after tons of work and tuning.
 +
 
 +
:[[Raven]]:
 +
:[[Raven]] is my most competitive 1on1 robot. It currently holds the 7th place in roborumble.
 +
 
 +
==Thoughts==
 +
:See [[/Thoughts on targeting]] for [[Targeting]].
 +
 
 +
==Formulas==
 +
:See [[/MEA]]
 +
 
 +
==Projects I am Working On==
 +
:A garbage free Neural Network library for future robots. I'd really like to set the standards Skilgannon and Rednaxela set with their kd-trees :)
 +
 
 +
 
 +
==Robot Names That You Can Use==
 +
*Parallax
 +
*Tempus
 +
*Inferno
 +
 
 +
==For Fun==
 +
;Idioms
 +
:Being SittingDuck against: Enemy's gun is very strong.
 +
:Usage: [[Walls]] is SittingDuck against [[Tomcat]].

Latest revision as of 09:52, 12 April 2021

Me

I am a Turkish student currently in high school. My teacher introduced me to Robocode and it probably is the only reason I do programming today.
Over the time, I've come to realize what a complicated optimization problem robocode was.
Since the July of 2020, I have spent more time learning other languages, most of them being lower level and I realized how inefficient most of my code was. Although I don't really feel like using Java anymore, I don't think I can possibly give up on robocode just yet. At least not before I give a final shot with Neural Networks ;)

Bots

I have made many robots, mostly experimental and most of them are retired.


ColdBreath:
My first competitive NN bot. Uses Roboneural. Reached a ranking of 55 with only 1 predictor for both movement and gun.
Oculus:
Oculus reached a ranking of 35 but was later removed temporarily since it was too slow.
WhiteFang:
WhiteFang is my third robot to be released. It's now in top 10 after tons of work and tuning.
Raven:
Raven is my most competitive 1on1 robot. It currently holds the 7th place in roborumble.

Thoughts

See /Thoughts on targeting for Targeting.

Formulas

See /MEA

Projects I am Working On

A garbage free Neural Network library for future robots. I'd really like to set the standards Skilgannon and Rednaxela set with their kd-trees :)


Robot Names That You Can Use

  • Parallax
  • Tempus
  • Inferno

For Fun

Idioms
Being SittingDuck against: Enemy's gun is very strong.
Usage: Walls is SittingDuck against Tomcat.