From: rars-d-request@lysator.liu.se Subject: rars-d Digest V97 #1 X-Loop: rars-d@lysator.liu.se X-Mailing-List: archive/volume97/1 Precedence: list MIME-Version: 1.0 Content-Type: multipart/digest; boundary="----------------------------" To: rars-d@lysator.liu.se Reply-To: rars@lysator.liu.se (Rars mailing-list) ------------------------------ Content-Type: text/plain rars-d Digest Volume 97 : Issue 1 Today's Topics: RE: Neural Net / GA Drivers RE: Car colours in RARS. Re: Car colours in RARS. Re: Neural Net / GA Drivers RE: Car colours in RARS. Re: Neural Net / GA Drivers RE: Car colours in RARS. RE: Car colours in RARS. Re: Neural Net / GA Drivers Re: Car colours in RARS. RE: Car colours in RARS. RE: Car colours in RARS. Announcing RARS 0.67 for Linux and OS/2 RE: Car colours in RARS. Administrivia: There has been some problems with the digest generation the last couple of months. The last digest was sent out August 8, but I didn't realize the problem until yesterday. Anyway, the previous digest can be found at http://www.lysator.liu.se/~oscar/rars/volume97/76.2 as it is 170k I thought that I would not send it out and it was easier to solve the problem by simply removing from the spool. If you want it sent to you, you can mail me on oscar@lysator.liu.se saying so. Sorry for any inconvenience this may have caused you. Oscar Gustafsson, oscar@lysator.liu.se RARS mailing-list administrator ------------------------------ Date: Tue, 4 Nov 1997 13:46:48 +0100 From: Torben Thellefsen To: "'Rars mailing-list'" Subject: RE: Neural Net / GA Drivers Message-ID: <01BCE928.2BCD4DD0@PC6> Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit [Robert Wilderspin wrote:] >Does anyone out there have experience with writing a RARS driver >using neural nets, and/or training one with a genetic algorithm? If you can read Dutch, try: http://carol.fwi.uva.nl/~stephanh/cursusnn/rars.html It's something about back-propagation or something (AI isn't my strong side :-). Regards, Torben ------------------------------ Date: Tue, 4 Nov 1997 13:43:42 +0100 From: Torben Thellefsen To: "'Rars mailing-list'" Subject: RE: Car colours in RARS. Message-ID: <01BCE928.27FE6B80@PC6> Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 8bit [Ralph Scott wrote:] >Laptops? Personally, I wouldn't plan for them. Someone can make >their own changes. I would say shoot for >=256 color scheme. >In javarars, I had each car return the color of its choice, so the >array didn't need to include the colors. I'd like to include this in the "normal" RARS as well - later. The 0.67 I'll release tomorrow will use the new way of defining car colours. If people complain about it, I'll change it back (maybe :-). >I agree with Henning that is is bad for portability, and agree with >you that for the most part, (ansi compatible compilers) it shouldn't >matter. Henning doesn't claim that compilers will barf on it, but that it will cause problems with environments/systems (eg. laptops) that don't have enough colours available and where it would be better to define a new set of colour pairs in gi.cpp instead of in carz.cpp. But since we have to edit carz.cpp anyway to include our own car(s), this doesn't matter (IMO). Slightly related: How many votes do I get for wanting to move the car definitions out of carz.cpp (and leave it truly platform independant) and into its own file? This will not be in 0.67, but later. Regards, Torben ------------------------------ Date: Tue, 04 Nov 1997 09:56:22 -0500 From: Ralph Scott To: Rars mailing-list Subject: Re: Car colours in RARS. Message-ID: <345F3796.4744@netusa.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Torben Thellefsen wrote: > The 0.67 I'll release tomorrow will use the new way of defining car > colours. If people complain about it, I'll change it back (maybe :-). Might as well state it now. Before you go through the trouble. > Henning doesn't claim that compilers will barf on it, but that it will cause problems with environments/systems (eg. laptops) that don't have enough colours available and where it would be better to define a new set of colour pairs in gi.cpp instead of in carz.cpp. Is anyone currently running it on a laptop? > But since we have to edit carz.cpp anyway to include our own car(s), this doesn't matter (IMO). > > Slightly related: > How many votes do I get for wanting to move the car definitions out of carz.cpp (and leave it truly platform independant) and into its own file? This will not be in 0.67, but later. This is fine with me. ---ralph ------------------------------ Date: Tue, 04 Nov 1997 10:10:03 -0500 From: Ralph Scott To: Rars mailing-list Subject: Re: Neural Net / GA Drivers Message-ID: <345F3ACB.7C5D@netusa.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Torben Thellefsen wrote: > > [Robert Wilderspin wrote:] > >Does anyone out there have experience with writing a RARS driver > >using neural nets, and/or training one with a genetic algorithm? > > If you can read Dutch, try: > http://carol.fwi.uva.nl/~stephanh/cursusnn/rars.html > > It's something about back-propagation or something (AI isn't my strong side :-). Or check out (in english)... http://www.css.tayloru.edu/~snelson/rars/index.html ---ralph ------------------------------ Date: Tue, 4 Nov 1997 16:19:10 +0100 From: Torben Thellefsen To: Rars mailing-list Subject: RE: Car colours in RARS. Message-ID: <01BCE93D.66957900@PC6> Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 8bit [Ralph Scott wrote:] >> The 0.67 I'll release tomorrow will use the new way of defining car > colours. If people complain about it, I'll change it back (maybe :-). >Might as well state it now. Before you go through the trouble. I already did. Check the first message in this thread (I think you've already replied to it once ;-), posted on October 29th. The change is already implemented (has been for months now, actually). >Is anyone currently running it on a laptop? Well, Henning uses one, but not exclusively, AFAIK. >> How many votes do I get for wanting to move the car definitions out of carz.cpp (and leave it truly platform independant) and into its own file? This will not be in 0.67, but later. >This is fine with me. That makes two of us . But I somehow expected you, as the race director, to favour this idea ;-) And I ask you (and the rest of the gang) again: Where should I announce the new release? Well, here, of course ;-) but in what newsgroups (if any)? Regards, Torben ------------------------------ Date: Tue, 4 Nov 1997 17:06:51 +0100 From: kim@ida.his.se (Kim Laurio) To: rars@lysator.liu.se Subject: Re: Neural Net / GA Drivers Message-Id: <199711041606.RAA08792@balder.ida.his.se> Rob wrote: > My project supervisor has informed me that a common approach is to use > GAs to get to a "good enough" approximation of the global optimum, and > then train further using an ANN to get it exact. I can't back this > up, but it's an idea to think about. Well, I'm not sure that I'm following you here but, isn't ANN just the architecture (i.e. no learning method specified)? I thought it was a common approach to use a GA to quickly get a decent solution and then use some sort of gradient based search (which is much faster in pinpointing the local maxima). For an alternative to GA (since you talked about comparing different learning methods instead of architectures), you could take a look at Evolution Strategies (ES). There was a short article on a variant of those in the April issue of Dr. Dobbs (I think?). It was about a method called "Differential Evolution". > One of the biggest problem I'm going to have, it seems, is selecting > the best data to optimise. For a simple car, good enough to get > around most tracks with no traffic, what would you suggest as a good > set of inputs? My current list is this: > > cur_rad, cur_len, nex_rad, nex_len, after_rad, after_len, v, vn, > to_lft, to_rgt, and to_end. Assuming that it is sufficient for your thesis to get a car to survive on most tracks (not necessarily being the fastest), one idea could be to do some pre-processing for the input to the ANN so that you give it multiple data of the same type. To understand what I'm thinking of you could imagine the car having a few range sensors that return the distance or/and relative speed to obstacles around it. Of course, the interesting case would be the driver that manages to navigate through traffic. It could be interesting to see the emerging global behaviour of many such drivers (if successful). This could have applications in traffic simulations when projecting new highways etc. Good luck! KL ------------------------------ Date: Tue, 4 Nov 1997 11:25:17 -0500 (EST) From: Daniel Brooks To: Rars mailing-list Subject: RE: Car colours in RARS. Message-ID: Content-Type: TEXT/PLAIN; charset=US-ASCII On Tue, 4 Nov 1997, Torben Thellefsen wrote: :)Slightly related: :)How many votes do I get for wanting to move the car definitions out of carz.cpp (and leave it truly platform independant) and into its own file? This will not be in 0.67, but later. I like this idea. I think we should have one file that has everything that is platform dependant, that way we only need to chnge one file to port it. TTYL, Daniel Brooks BTW, is there a port for monochrome monitors? I haven't looked because I don't expect to find one. ________________________________________________________________________ __ Daniel Brooks - d-brooks@usa.net, dbrooks@geocities.com / /\ http://www.geocities.com/SiliconValley/Vista/2083 / / \ / / /\ \ HANDLE WITH EXTREME CARE: This Product Contains Minute / / /\ \ \ Electrically Charged Particles Moving at Velocities in / / / \ \ \ Excess of Five Hundred Million Miles Per Hour. / /_/____\ \ \ /__________\ \ \ "We're very sorry, Mister Schrodinger, \_____________\/ but the cat refuses to go in the box." ------------------------------ Date: Tue, 4 Nov 1997 17:29:23 +0100 (MET) From: Maido Remm To: Rars mailing-list Subject: RE: Car colours in RARS. Message-ID: Content-Type: TEXT/PLAIN; charset=US-ASCII On Tue, 4 Nov 1997, Torben Thellefsen wrote: > > >> How many votes do I get for wanting to move the car definitions > >> out of carz.cpp (and leave it truly platform independant) and into > >> its own file? This will not be in 0.67, but later. > > >[Ralph:]This is fine with me. > > That makes two of us . > I think it is nice to have colors defined together with driver-program names in carz.cpp like you planned for now. Time shows whether we will like it or somebody will have problems. For splitting files I would think twice - this should be done only once and done in most efficient way. It would be nice to have less files. One possible way to do that: * one OS-dependent file could include gi.cpp and os.cpp * draw.cpp (track length calculation taken out of it) * movie.cpp (carz.cpp + misc.cpp + report.cpp + track.cpp) could be done in two files * a) car and its physics, track (we dont change this often, do we?) * b) race manager, report manager, screen manager etc. (main program?) So we would have 5 files instead of 8. This is just my imagination. I have no idea whether this system is better functionally or regarding to portability but it looks simpler to me... Maido ------------------------------ Date: Wed, 5 Nov 1997 08:57:23 +0100 (MET) From: bm321465@muenchen.org (Henning Klaskala) To: rars@lysator.liu.se Subject: Re: Neural Net / GA Drivers Message-Id: <199711050757.IAA00856@hk.muenchen.org> Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Maido Remm wrote: > > On Tue, 4 Nov 1997, Robert Wilderspin wrote: > > > >Also consider of doing it in Hennings way - just try each point in > > >each segment independently and see what happens. This actually depends a > > >lot which values you are optimizing and how independent they are to each > > >other. Usually my optimizer mutates several adjacent points at once and then runs a few laps to see whether the mutation was an improvement. > > Henning's approach, from what you've written there, sounds like > > optimising it by hand. Is that right? I did that with my midget car. > > It worked very well for the track I did it on, but I have no way of > > knowing whether it was a local optimum or not. > > > No I dont think a programmer should do anything 'by hand' :-). > I think he simply tries all possibilities consequently and quits if car is > getting worse, then takes next segment and tries that. No, I don't do it segment-by-segment. For each mutation I coose a random number of adjacent points in one or more adjacent random segments. Also I can't try all possibilities because there are too many. > His car is using > extremely simple driving alghoritm and this makes it efficient and > probably easy to optimize. He uses optimized speed values as well, Probably true. I just don't know whether the simple driving algorithm makes it easier to optimize. It still takes days to get close to the (local?) optimum. Henning ------------------------------ Date: Wed, 5 Nov 1997 08:57:16 +0100 (MET) From: bm321465@muenchen.org (Henning Klaskala) To: rars@lysator.liu.se Cc: rscott@NetUSA.Net Subject: Re: Car colours in RARS. Message-Id: <199711050757.IAA00849@hk.muenchen.org> Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Ralph Scott wrote: > > Torben Thellefsen wrote: > > > Henning doesn't claim that compilers will barf on it, but that it > > will cause problems with environments/systems (eg. laptops) that > > don't have enough colours available and where it would be better > > to define a new set of colour pairs in gi.cpp instead of in carz.cpp. Meanwhile I solved this problem in gi.cpp for X11 (UNIX): I added a function that automatically changes the car colors if several cars have the same colors. So I don't have to change the colors manually in the "platform independent" file carz.cpp to adapt it to my platform. (Several cars had been painted black when there were not enough colors available.) > > Is anyone currently running it on a laptop? I developed all my drivers and the track generator on an old 386 laptop. The monochrome LCD screen caused no problems because the laptop is connected to a color monitor. :-) The problem was not the laptop, but the 16 color mode that I had to use. This could be a problem on any outdated hardware, not just laptops. > > How many votes do I get for wanting to move the car definitions out > > of carz.cpp (and leave it truly platform independant) and into its > > own file? This will not be in 0.67, but later. > > This is fine with me. With me, too. Henning ------------------------------ Date: Wed, 5 Nov 1997 09:22:22 +0100 From: Torben Thellefsen To: "'Rars mailing-list'" Subject: RE: Car colours in RARS. Message-ID: <01BCE9CD.665BA1E0@PC6> Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 8bit [Maido Remm wrote:] >I think it is nice to have colors defined together with driver-program >names in carz.cpp like you planned for now. Time shows whether we will >like it or somebody will have problems. Hopefully noone (has problems, that is) :-) >For splitting files I would think >twice - this should be done only once and done in most efficient way. It >would be nice to have less files. One possible way to do that: Ah yes. >* one OS-dependent file could include gi.cpp and os.cpp >* draw.cpp (track length calculation taken out of it) >* movie.cpp (carz.cpp + misc.cpp + report.cpp + track.cpp) could be done in two files > * a) car and its physics, track (we dont change this often, do we?) > * b) race manager, report manager, screen manager etc. (main program?) >So we would have 5 files instead of 8. >This is just my imagination. I have no idea whether this system is better >functionally or regarding to portability but it looks simpler to me... It certainly has the potential for being easier portable. In making 0.67, I tried to make the Linux and OS/2 versions as similar as possible. Well, it was virtually impossible because the OS/2 version (0.66 based) does the UI in a different way and has a track class. These changes are implemented all over the code, so I couldn't easily find the changes and revert them to the 0.65 way that the Linux 0.67 is based upon. So the OS/2 version remains different (for now) and most fast cars will not work with it (unless someone converts them to use the track class). So reorganizing the files is an idea that I'm quite willing to support :-) And I agree with you that we should think twice (or thrice) before doing it - I'm going on a five day vacation now and will try to find some time to think about it (I'll of course bring a copy of your mail for inspiration :-). Here are some (quick) thoughts: - main() should be as platform independant as possible, calling platform code where applicable. - we must remember that at least the OS/2 version has a different UI (where the panels are moved out of the main window - I like it). - we must make sure that existing drivers work without problems. More when I return. Regards, Torben ------------------------------ Date: Wed, 5 Nov 1997 09:28:24 +0100 From: Torben Thellefsen To: "'Rars mailing-list'" Subject: RE: Car colours in RARS. Message-ID: <01BCE9CD.6A8DD9E0@PC6> Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 8bit [Daniel Brooks wrote:] >:)How many votes do I get for wanting to move the car definitions out of carz.cpp (and leave it truly platform independant) and into its own file? This will not be in 0.67, but later. >I like this idea. I think we should have one file that has everything that >is platform dependant, that way we only need to chnge one file to port it. Well, to port it to a new platform you'll need to change more than one file. But to include your own driver you change one file, which does not contain other code than car definitions. See also Maido's suggestion about reducing the number of source files. >BTW, is there a port for monochrome monitors? I haven't looked because I >don't expect to find one. It would be extremely hard to tell the drivers apart ;-) Unless the port was written to use different fillpatterns instead of colours... And to answer your question: No, never heard of one. Not even grayscale. Regards, Torben ------------------------------ Date: Wed, 5 Nov 1997 13:26:21 +0100 From: Torben Thellefsen To: "'Rars mailing-list'" Subject: Announcing RARS 0.67 for Linux and OS/2 Message-ID: <01BCE9EE.6B489B60@PC6> Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Announcing RARS 0.67 for Linux and OS/2 ======================================= I have uploaded to ftp.ijs.com (in /incoming) the following files: rars-linux-0.67.tar.gz - The Linux version rars-os2-0.67.zip - The OS/2 version Please download and test whichever version(s) is relevant to you :-) Version 0.67 includes the pitting code from Ralph. This has also been changed in all the included drivers. Both platforms now have similar makefiles for easier recompilation. The Linux version includes code from Henning Klaskala to run races on a random track (specify 'random' as the track name). I was unable to implement this in the OS/2 version for now (see below). Henning also contributed with some performance improvements for the floodfill code on Linux/X (works now and is quite fast). It was (and remains) my intention to build the Linux and OS/2 versions of RARS from the same set of platform independent source files. Unfortunately, this was not possible within the time I had available for this release, so the OS/2 version remains different on these items: - get_track_description() returns a class instead of the track_desc structure expected by most really fast cars (who consequently won't work). - The automatic generation of a random track isn't implemented. - The platform independent files are not. This will be changed in future versions (yes, I intend to make future versions). Comments / bugreports / bugfixes / suggestions / ideas can be mailed to torben@dir.dk or (in case of general things) discussed on the mailing list. I'll be on vacation the next 5 days and thus be unavailable for comment, but feel free to mail me anyway ;-) This release will not be announced on any newsgroups (not by me, at least). I want a more easily portable version before I do that. Although it may seem so to some, I'm not bashing the OS/2 porters for their work - after all, it was the availability of an OS/2 version that got me into this game (perhaps some might want to bash the OS/2 porters for *that* reason ;-). Trying to keep RARS rolling, Torben Thellefsen torben@dir.dk ------------------------------ Date: Wed, 5 Nov 1997 07:37:34 -0500 (EST) From: Daniel Brooks To: Rars mailing-list Subject: RE: Car colours in RARS. Message-ID: Content-Type: TEXT/PLAIN; charset=US-ASCII On Wed, 5 Nov 1997, Torben Thellefsen wrote: :)[Daniel Brooks wrote:] :) :)>:)How many votes do I get for wanting to move the car definitions out of carz.cpp (and leave it truly platform independant) and into its own file? This will not be in 0.67, but later. :) :)>I like this idea. I think we should have one file that has everything that :)>is platform dependant, that way we only need to chnge one file to port it. :) :)Well, to port it to a new platform you'll need to change more than one file. But to include your own driver you change one file, which does not contain other code than car definitions. :)See also Maido's suggestion about reducing the number of source files. :) :)>BTW, is there a port for monochrome monitors? I haven't looked because I :)>don't expect to find one. :) :)It would be extremely hard to tell the drivers apart ;-) Yea. That is the one gripe I have with my mono monitor. :)Unless the port was written to use different fillpatterns instead of colours... :) :)And to answer your question: No, never heard of one. Not even grayscale. I might be making one. I just need to show one car (mine). There is also a problem with the source I have. It freezes my 386, even when I run it in no display mode. TTYL, Daniel Brooks PS: I just had a brainstorm. How bout adding a third mailing list that was dedicated to mailing the source for RARS? That way every time we geet a new version, we can put it on the mailer to distribute. it could probably also serve it up on request. :) :) :)Regards, :)Torben :) :) :) ________________________________________________________________________ __ Daniel Brooks - d-brooks@usa.net, dbrooks@geocities.com / /\ http://www.geocities.com/SiliconValley/Vista/2083 / / \ / / /\ \ HANDLE WITH EXTREME CARE: This Product Contains Minute / / /\ \ \ Electrically Charged Particles Moving at Velocities in / / / \ \ \ Excess of Five Hundred Million Miles Per Hour. / /_/____\ \ \ /__________\ \ \ "We're very sorry, Mister Schrodinger, \_____________\/ but the cat refuses to go in the box." -------------------------------- End of rars-d Digest V97 Issue #1 *********************************