unforum

[Mikeyj]

I have absolutely nothing to add to this thread whatsoever, other than to say it really pleases me to see people happily playing with a toy

[BrianEnigma]

I have a new release of the ROVR application. Same as before--launch maze.jar, then rovr-1.1.jar. Changes include:

• Rudimentary mapping (yes, you can easily confuse it right now)
  
• "Magnet" mode that moves the ball exactly one square when you hit a movement key
  
• Speed selection
  
• Arrow keys work as well as the letter keys

The mapping will get confused if you stop half-way between squares. At this stage, it is only terribly useful for straight runs. The "magnet" mode stops exactly where you want it to about 95% of the time. Every once in a while, it ends up a pixel off, making it difficult to take a side passage.

Screen shot:


Jar file link:
http://perplexcitycardcatalog.com/notes/maze/rovr-1.1.jar

Source code link:
http://perplexcitycardcatalog.com/notes/maze/rovr-1.1-src.tgz

[Seej]

Well, I guess my concerns about brute forcing and reading the classes this for answers weren't particularly justified, though it would be nice to actually play the game by waving a laptop like a loon. If only I was sober enought to telnet into my onw box and make this work...........

[NovacaineX]

Well after working for a few hours on maze6 i have come to a dead halt. I cannot find anymore ways to go. Here's an image of the maze i got.

maze 6

EDIT: green is the start

[step]

I updated my program to keep track of the current relative postion.

It doesn't have all the nice extra functionality of Brian's, but if anybody wants it, it's here

The position is indicated on a button on the right of the status. if you click it, it will reset the position indicator back to 0,0. however, it doesn't actually move you; this is in case you get lost and have to reset the maze itself or when you switch to the next level.

EDIT: I updated the program again. added a "Step" button to toggle single square movement. Also major improvement in responsiveness

All the notes in my previous post remain in effect.

[Spankit]

Is it just me, or does maze 6 have some sort of hidden switch mechanism(s)? Certain paths do not open up until I've hit all the invisible walls, or so it seems to me. Maybe I'm just completely wrong and/or misinterpreting Brienigma's mapping feature on rovr-1.1 (a nice touch, by the way).

Also, just so you all don't think I'm completely crazy, I'm talking about the new maze.jar file that's been put out.

[BrianEnigma]

[Spankit]

[SteveC]

Well, my app is as advanced as it's been just for PC users.. Now available at:

http://www.kallisti.co.uk/Mapper.zip (same as others, just run maze.jar, then my creatively named WindowsApplication1.exe). It should have all the bits it needs to run with it now...

Feedback welcome by pm or here..

Notes: 1/ the magnet isn't perfect, you can miss side corridors.
2/ If you go diagonally across a space the map will get confused.

I can't resolve BOTH of those situations I'm afraid, so be careful out there kids...

Oh, and reset doesn't reset the maze, just my app..

[RobPal]

I agree with Spankit. There does seem to be some switch mechanism that opens and closes paths when you hit certain walls. I've been playing the game on a PC using BriEnigma's app with the magnet switched on. I can't see how it could get confused as with mazes 1 to 5, it worked fine. There's definitely something weird going on with maze 6.

[SteveC]

OK OK OK 1/ I'm with Brian, I've seen no evidence for the on again off again gates...

2/ I've updated http://www.kallisti.co.uk/Mapper.zip so that it's got a set of "feather/search" buttons. Two for each direction that just nudge the ball a little to search for mid-corridor exits.

Notes.

1/ Be patient, it assumes you'll wait for the ball to come to a standstill, DOESN'T force you to. If you change something while the ball is still moving, it'll probably corrupt your map.

2/ It's got a "unsafe" mode where you can move diagonally. Don't use it If you hit an open space it'll corrupt the map.

Other than that, post your maps once you've done them all

[Spankit]

Alright, so I spent about an hour this morning trying to figure out maze 6 before I got extremely bored and quit. I've attached the map that I got from Brienigma's program. There were definitely paths that did not open up until after I hit certain walls. Actually, I'm not sure if this map will be of any use seeing as how I got a different maze 6 when I decided to give it another try.

[BrianEnigma]

I posted this to IRC and figured I would put it here, too, for the sake of completeness. It is mainly of interest for the programmers that are working on mapping and autopiloting, since it is just source code and needs a basic C compiler (GCC, which is free, works just fine).

http://perplexcitycardcatalog.com/notes/maze/mapmaker.c

It allows you to create your own maps for the first 5 mazes. This way, you can set up your own conditions (side corridors, walls, T-intersections, turns, etc.) for testing your programs.

Something that may be of interest to others is that you can expand maze.jar, copy the file "e" in the root folder several times to replace "a," "b," "c," and "d" (but don't touch any of the *.class files), then re-zip it (keeping the directory structure intact), and you'll have a quick and easy way of going directly to maze 6 without having to navigate 1-5 because maze "e" (or maze 5) is just a straightaway. I am really not sure if this is "within the rules," or skips past anything important, so caveat emptor.

[BrianEnigma]

Also, 1.2 of ROVR, including:

• The concept of "leaning," which adds a mild (value 15) or strong (value 20) force to the right or left of your direction of travel--useful for finding side corridors, but totally blows away auto-mapping.
  
• Autopilot, which moves in the cardinal directions (no diagonals or leaning), a step at a time. Every time it hits a wall, it tries to make a right turn. This doesn't always work, as sometimes the wall hit makes it a pixel or two off from the turn's path, causing it to hit the wall and turn right again (backtracking it's path.) If it gets stuck (no movement for 2 seconds), it picks a random direction to move in.

I have tried various automatic movement (both with and without mapping) using diagonals. This includes a pure diagonal (+/- 15 on both axes), as well as a directional bias (20 in the main direction, 15 off to a side). This works fine for side passages, but when you come to a turn or "T" at the end of a corridor, it often gets stuck--hitting the wall straight-on often causes it to bounce back far enough to not be lined up for the turn.

Anyway, jar file for 1.2:
http://perplexcitycardcatalog.com/notes/maze/rovr-1.2.jar
Source code for 1.2:
http://perplexcitycardcatalog.com/notes/maze/rovr-1.2-src.tgz

Yes, the autopilot's brain is REALLY DUMB. What do you expect for the first iteration?

Does anyone have any thoughts on how to best automatically navigate the maps? It *HAS* to have something to do with diagonals, otherwise you would completely miss side corridors. Man...this would be so much easier if the input, output, and movement were nice digital steps rather than this analog movement.

[SteveC]

My prog aids in finding side corridors by hitting a diagonal for a set amount of time (I find 150ms best). This means it doesn't go into a new square, but will usually (or often) find a side corridor in that direction. The problem is that if you hold diagonal, you can go across an open space (four squares forming a larger square, or more), and that's almost impossible to map as it's difficult to tell it has happened!

Only thing I'd point out is that manually map navigating might be faster

I'm also very suprised that no-one has found an "exit" yet.. Leads me to believe we're barking up the wrong tree.