//----------------------------------------------------------------------------
//
// RARS Robot racing code DougE1.CPP from Doug Eleveld
//
// Developed for RARS 0.72 with DJGPP/Allegro
//
//----------------------------------------------------------------------------
//
// !!! DougE1.cpp must be associated with data file DougE1.dat !!!
// !!! otherwise nothing will work !!!
//
// Anybody can do whatever they want with this code, but if you use ideas
// I'd like you to send me an email.  This source code is public.
//
// The car 'recognizes' the road that it on by looking at the length, radius
// etc of the segments and looks up in a big table of optimized values that
// is in the datafile.  The optimizer is another program and I won't include
// it here as there is just simply too much code.  If it gets put on a track
// for which it is not optimized, the 'best fit' of the optimized values is
// chosen.  This usually means that it will crash alot on tracks for which it
// is not optimized, and probably won't make it even once around.  It is
// optimized for all the tracks the RARS 0.72 except, of course, for
// RANDOM.TRK.  Optimization uses a genetic algorithm with an initial path
// and speed that are set by hand.  I have not done anything to increase
// performance on un-optimized tracks, although I may do leave-one-out
// cross validation to find the best weights for 'recognizing' each segment.
// I am willing to share the code with pretty much anyone who asks.
//
// The car's path is dependant on the road, and not on global coordinates.
// This means that is is fairly stable, although it performs poorly
// (especially passing) when there are many short (compared to car length)
// sections of road like in MONACO.TRK or high speed straights that end in
// sharp curves like in AUSTRIA.TRK.  It loves long sweeping curves like
// SPA.TRK.  It performs well on MONZA-76.TRK for most of the track, but
// gets disturbed by the short chicane-thing and that destroys it's time.
// I have been thinking about an 'override' to global coordinates for these
// difficult sections, but I not sure if I can do that...
//
// What I have also implemented that I have not earlier seen here is that
// different approach speeds are used for different cars and anti-bulldozing
// code.  The approach speed of a car depends on how long DougE1 has been
// trying to pass them.  This is intended to avoid getting stuck behind
// someone that I cannot pass, which especially happens on short ovals.
// The passing code is only mediocre, even though it is quite complex.  It
// is particularly poor at passing groups of cars, therefore it doesn't
// usually do to well on the short ovals.  Sometimes it passes great, but
// when it does crash it usually crashes in a spectacular fashion.
//
// You may notice some wierd or non-optimal things in the physics here.  This
// is because I know very little about the physics of car racing (I'm an
// electrical engineer), but I do know a fair bit about optimization.  And
// that is what DougE1 is pretty much completely based on, optimized paths
// for each track.
//
// I got the inspriration for my pattern based matching stuff from looking
// at the Wappucar source, although I implemented it in a completely
// different way.  There is also a tiny little bit of code from Apex8 in the
// collision avoidance code.  Other than that I think it's a pretty original
// robot.
//
// I'd love to discuss the code with anyone, and would actually like to team
// up with someone who knows how to white a better collision avoidance than
// I have.  I'd might want to make one using global coordinates, based on the
// TURTLE9 source.  Or maybe make a car only for ovals with seperate fast
// path and passing path.