ICFP 2003 Programming Contest
Main Entry

Team Name:                "TeamSpud"
Memebers:                 Matt Quail
URL:                      http://madbean.com/icfp2003/
Email:                    spud[at]madbean[dot]com
Implementation Language:  Java (JDK1.4)


Submission Contents
-------------------

- Trace files for all the (non-rally) courses:
    1_Simple.trc
    2_Hairpins.trc
    3_Sepang.trc
    4_EatYouAlive.trc
    5_Car.trc
    6_IcfpContest.trc
    7_Gothenburg.trc
    8_ManyWays.trc
    9_PhilAndSimon.trc
- README.txt    This filefile.
- teamspud.jar  A Java program to run the GUI
- build.xml     Ant build file
- src/*         The Java source code.
- Tracks/*      The competition tracks
- guides/*      Guides used by the drivers/GUI


Trace results
-------------
 track 1 steps=11434
 track 2 steps=17537
 track 3 steps=20833
 track 4 steps=22287
 track 5 steps=6733
 track 6 steps=6547
 track 7 steps=6945
 track 8 steps=11575
 track 9 steps=21481


Running the program
-------------------

You can run a GUI simulator by typing:

$ java -jar teamspud.jar

You can add/insert/move/delete way-points by hitting the appropriate
button. (You can load and save guidepoints too).

You can run my "driver algorithm" by hitting the "Run Driver" button,
which attempts to follow the guidepoints.

You can load a .trc file from the "file" menu. Then hit the "Run
Trace" button.

During the simulation, the car will leave behind a blue line. Dark
blue means slow, white means fast. Use the "look" tool to determine
the actual speed at a particular point (click with the mouse, the
speed is shown in the status bar).


Description
-----------

This program implements a "driver" algorithm. The driver needs to be
given a set of way-points. It pre-computes a set of "target speeds"
for each waypoint, then starts driving!

This driver prefers to roll to a target speed (as opposed to
accelerating then braking), so it is often necessary to add way-points
that act as a "start-brake-here" point on a long straight.

To begin with, I couldn't determine a process that would compute a
set of line segments from start-to-finish (without intersecting any
walls). I couldn't grok the "topology" problem. So, that is why I have
implemented a "manual waypoint" approach. You must manually set up the
"topology", then the program takes it from there.

In the last 12 hours, I did come up with an algorithm that could solve
the "topology" problem, but didn't have the energy/time to implement it.

This has been fun. Congrats to the winners!

=Matt

PS: "9_PhilAndSimon.trc". What were you thinking!?