Difference between revisions of "User:CrazyBassoonist"
(A good robot, finally) |
|||
Line 18: | Line 18: | ||
February 16, 2009: Made my first learning gun... I'm not sure what type it would be, maybe a non-virtual gun. What it does is it assigns each of two different gun types random hit percentages at the beginning of a match and then uses them each randomly until it has enough data to choose one.However, because it has no certain way of knowing which gun fired the bullet that hit the other robot, it is heavily biased towards whatever gun it chooses at the beginning, thus leading to its extremely random score over short periods of time. Without movement added it gets 74% on average against walls over a thousand rounds, and anywhere from 81% to 38% over 35 rounds. | February 16, 2009: Made my first learning gun... I'm not sure what type it would be, maybe a non-virtual gun. What it does is it assigns each of two different gun types random hit percentages at the beginning of a match and then uses them each randomly until it has enough data to choose one.However, because it has no certain way of knowing which gun fired the bullet that hit the other robot, it is heavily biased towards whatever gun it chooses at the beginning, thus leading to its extremely random score over short periods of time. Without movement added it gets 74% on average against walls over a thousand rounds, and anywhere from 81% to 38% over 35 rounds. | ||
− | May | + | May 03, 2009: Made my first decent gun. Its what it does is it draws a bunch of dots around the enemy robot. When the enemy robot bumps into one, it records the point it went to and the point it was coming from. Then it basically works like a single tick pattern matcher, using the point that the enemy most often travels to from the one its at and then recreating its predicted movement from that. I daresay it is beautiful to watch, especially since it paints it's enemy's predicted movement. |
+ | May 09, 2009: What a day! Today I made a GF gun segmented on enemy distance and velocity from scratch. I finally figured out how to use multiple classes, and In addition to the GF Gun I also made a virtual gun system including the GF gun, a head-on targeter, a circular targeter, and a segmented averaged velocity circular targeter. I put this onto a weird adaptive movement I made a few days ago, and voila! My first good robot. | ||
== My Projects == | == My Projects == | ||
[[Super Sample Bots]]- A set of bots that are based on the sample bots. They are intended to be the next challenge for beginners after they have a robot that can beat all of the regular sample bots. | [[Super Sample Bots]]- A set of bots that are based on the sample bots. They are intended to be the next challenge for beginners after they have a robot that can beat all of the regular sample bots. | ||
[[Category:Bot Authors|CrazyBassoonist]] | [[Category:Bot Authors|CrazyBassoonist]] |
Revision as of 23:47, 9 May 2009
Contents
About Me
I learned about Robocode while doing a report on artificial intelligence for school, and so far I've found it very enjoyable. Unfortunately, I'm extremely new (as in no previous experience) to Robocode, Java, and programming in general. I am also not yet old enough to be taking geometry, which seems to be important for this game. However, I'm a fast learner and I'm sure I can figure it out. As you may have guessed, one of my hobbies is playing the bassoon.
My Robots
DolphinAR- Uses stop and go movement with a learning gun.
My Progress
February 16, 2009: Made my first learning gun... I'm not sure what type it would be, maybe a non-virtual gun. What it does is it assigns each of two different gun types random hit percentages at the beginning of a match and then uses them each randomly until it has enough data to choose one.However, because it has no certain way of knowing which gun fired the bullet that hit the other robot, it is heavily biased towards whatever gun it chooses at the beginning, thus leading to its extremely random score over short periods of time. Without movement added it gets 74% on average against walls over a thousand rounds, and anywhere from 81% to 38% over 35 rounds.
May 03, 2009: Made my first decent gun. Its what it does is it draws a bunch of dots around the enemy robot. When the enemy robot bumps into one, it records the point it went to and the point it was coming from. Then it basically works like a single tick pattern matcher, using the point that the enemy most often travels to from the one its at and then recreating its predicted movement from that. I daresay it is beautiful to watch, especially since it paints it's enemy's predicted movement.
May 09, 2009: What a day! Today I made a GF gun segmented on enemy distance and velocity from scratch. I finally figured out how to use multiple classes, and In addition to the GF Gun I also made a virtual gun system including the GF gun, a head-on targeter, a circular targeter, and a segmented averaged velocity circular targeter. I put this onto a weird adaptive movement I made a few days ago, and voila! My first good robot.
My Projects
Super Sample Bots- A set of bots that are based on the sample bots. They are intended to be the next challenge for beginners after they have a robot that can beat all of the regular sample bots.