[Home]Nathanael

Robo Home | Changes | Preferences | AllPages

My Robots

None are particularly good (in fact they're darn near abysmal), because at the moment I have neither the time nor the patience to implement decent targeting or movement algorithms.

Hardware

1x AMD Athlon XP 2100+/512MB RAM (main desktop/development workstation)

1x Intel Celeron 600/128MB RAM (Internet proxy, Bluetooth gateway and CVS server)

Software

[SuSE] Linux 9.x! The best Linux distro out there, especially for Java development. It comes with Sun's JRE and J2SDK, Ant, Tomcat, JBoss (not that I use JBoss; I use Sun app Server 8) etc and SuSE? 9.1 comes with kernel 2.6, which speeds up Java a lot with the Native Posix Threads Library.

Windows XP Home, because I have to.

Development tools

vim - The best text editor, with autoindent, syntax highlighting, tab expansion etc etc.

ant - Build tool for Java. Makes things heaps easier to work with. No more "javac -classpath ~/robocode/robocode.jar src/blah/blah/blah". One command - "ant"

Sun's J2SDK 1.4.2

Why Robocode?

For my course at Uni (Software Engineering at [Curtin University of Technology]) one of our units has a semester project designed to teach us software requirements, project management etc and to do that, we build a Robocode robot. The details of it are secret, of course, and will be until the competition date. We (my team of four) have named it Sigterm - the UNIX interprocess signal for "Excuse me, Mr Process. Would you be so kind as to die gracefully?"


Welcome to the robowiki! -- PEZ

Welcome to the robowiki. Hopefully the competion around here is not as tough as you will face at school. I for one would not mind following along to see how your bot ends up placing. Please keep us up to date. -- jim

I will as soon as I can, which unfortunately won't be for a while yet. Early test robots are looking quite promising. -- Nathanael

Small update. Because we have to do this in teams with project management, configuration management etc, we split up the work into four areas (can't tell you what they are yet). So far two areas have been fully implemented, and still waiting on the other two modules. The CodeSize for the two implemented modules, along with a testing robot each come to a surprising 5752 together. -- Nathanael

That does not suprise me. Look at te size of a bot like SandboxDT. It is large. And with yours being a university project, you probably have all the requirements for debugging, maybe JUnits tests, etc. That size seems about right to me for half a bot. I am still very interested to see how this turns out. I can not imagine applying project management and having different groups working on different pieces of the code.-- jim

Large? U definitely donīt know how large can be a code! SS & SF are about 250,000 bytes... Full of crap, commented & unused code in itīs guts... Nothing that i would recommend for anybody, of course... -- Axe

Welcome, Nathanael. I'm looking forward to the SIG bots, it's been a while since anyone made a series of bots. My Duelist series, Kawigi's Flood line, and jam's Raiko bots are the only ones that come to mind.

@Axe I think you're confusing class size with CodeSize. SilverFist 1.1WT has a CodeSize of 43885. However, it's worth noting that it has the largest CodeSize of any bot in the RoboRumble@Home, other than Homer.Barney which I don't count since it is only huge because the author packaged a giant statistical library into it. --David Alves

Odd not to count Barney I think. SS and SF are big for about the same reasons as Barney actually. They carry around libraries of unused stuff. I sometimes figure on what on earth bots like DT (36539), Shadow (21461) and PrairieWolf (23780) are carrying around. Nathanael's 5752 for two modules could maybe be compared to CassiusClay .00b2 which weighs in at 5917. And that's including full GL debug capability. -- PEZ

The difference is that Barney is huge because he's including code that he didn't write, whereas Axe wrote all of the code in SilverSurfer, Paul wrote all of the code in DT. So, Axe's bots have the most *original* code. --David Alves

"None are particularly good"... Better than my first attempts. ;-) --David Alves

Sigterm 1.0 is released! Didn't do that badly in the uni competition. The other bots were pretty good, but all were fairly close. The first trial melees were failures (coming last in a 34 bot melee) because of a bug in the Windows version of Java 1.4.2 which occasionally doesn't allow saving. But we fixed that and a few other things and did quite well in the final competition held today (I can't remember actual scores, except that it came 3rd, 4th, can't remember and 13th for different battles of varying numbers of participants). Seth was one of the competitors. Final codesize was 11797, with about 2200 LOC written. The code wasn't being marked, so some of the code in the other sections (not in the movement) is rather sloppy. If you want and can catch my server online, the CVS tree is available at http://glacier.widge.org/cgi-bin/cvs/viewcvs.cgi/. -- Nathanael

The targeting in Sigterm is weird though. The pattern matcher is broken and disabled, probably because of a bad waves implementation. An earlier Sigterm testbot called FireTestBot? used random guess factor targeting and could beat Rapture 50:50 on points and rounds. But targeting can get confusing, so I don't think that Sigterm is that good. On some things FireTestBot? is better than Sigterm, sometimes Sigterm rather than FireTestBot?. -- Nathanael


Robo Home | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited November 5, 2004 10:06 EST by Nathanael (diff)
Search: