12 seconds to create the fast spline is pretty bloody impressive. So much easier than trying to drive that entire perfect lap. Just awesome!!!!!!! Sent from my GT-I9505 using Tapatalk
Well, I think this is unacceptable. Your script should do the AIW path instantly, not in some seconds.. Keep up the good work!
Dearest Tohtori Jokeri, Sincerest apologies for the decidedly unremarkable bit of coding that resulted in your witty post. Upon reading it, I called over to Bink Aviation, had my jet readied, and immediately flew to Finland to meet with the Distinguished Fellowship of Computer Scientists at the University of Helsinki (pictured below). After discussing your post with the Fellows for several hours, I came to two striking conclusions: 1. One of the Fellows isn't even a fellow. 2. It should be possible to write some code that will complete the Fast Line three full days before you even start building the track. Hopefully, this new benchmark of -72 hrs (negative seventy two hours) will quell the unrest that has overcome your fertile township, and help you to persuade the locals to put down their pitch-forks & torches, and get back to full production of the dental floss that has, for seven centuries now, earned them respect as the Jewel of the Frozen Tundra. Thanks again for your post, and always remember, when in doubt... turn left. View attachment 17773
Any updates Bink? Got some Annual Leave coming up soon would be awesome to get some AIW's done hopefully with the use of your tool Sent from my GT-I9505 using Tapatalk
Hey Butters.... After I got the prototype of the vector calculus function for the fastGen working a few weeks ago, I've been working to formalize it into it's own data structure. This led to 5 ideas on how to perfect the fast line generator.... the first 3 ideas use argument permutations of the new generic vecCalc functions... the other two ideas are ways to more accurately pick corner apex and turn in / out points. There was an issue with very short shallow corner apex finding (which also affected chicanes).... and another with turn in / out points on corners close (enough) to 180 degrees. Only have the math on those at this point (on paper), but it's pretty solid, and should help accuracy on shallow apices, and corners tighter that 120 degrees... never been too happy with the results on that stuff. Should be able to handle them perfectly... soon. To more broadly implement the new vecCalc structure's functions, there were some major infrastructure changes that needed to be made to the fastGen (data that needed to be in place). It was kind of traumatic ripping the heart out of the fastGen, as it was fast, and pretty accurate at about 98% of the things it encountered on tracks. I'm hoping these changes will take care of those last few tough things to crunch. [... one of which was very slightly curving straight-aways ...] So... the first week of September was formalizing the (new) vector calculus data structure, .... second week was restructuring the fastGen, ... and this past week has been trying to get them to 'play nicely' with each other. Next week will be plugging the new generic vecCalc functions into the new 'slots' in the fastGen decision tree. There were some 'holes' in the coverage of possibilities in the decision matrix of the fastGen that now should be covered. Situation awareness. That is to say... when finished, the huge number of things that could occur when plotting a fast line should now be covered. It's a matter of smoothly connecting the range of "where did the car just come from" ... to the ... "where does it need to be next". The new fastGen infrastructure should be able to do this ... again ... when I am finished plugging vecCalc functions with the proper arguments into it. This new matrix ends up handling a vast number of track architecture possibilities. I'm hoping to have this fastGen stuff done by the first week in October... when I can finally get to the implementation of the new fast line 'vewport follow' simulation (in the spline edit struct) that I came up with a couple months ago (currently a couple of tested prototype functions). Then things should be ready again for a few weeks of general testing (on all those test tracks I wrote about a few posts ago). Most everything else has gone through about 3 revisions so far. When I first started this, it was a matter of getting certain things written... or prototyped ... so that other things that depended on them could be written. When some of it started to work perfectly, other functions that just worked 'OK' became pretty obvious as deficiencies. As you might guess.... it's waay easier to come up with solutions to something once you can 'see' it in action, and get a feel for how it operates. That's where I'm at ATM.... ~33,000 lines of code.... (v191 a little smaller)... Going to get back at it in a couple hours after I watch the Singapore GP.... [Release the KRaikken ! ]
Sometimes is better to disconect. We'll be here, waiting for your news. Thanks bink. Enviado desde mi E2003 mediante Tapatalk
Back. Wrote a little script to quickly make tracks ... to test detecting and fitting a fast line for the various general corner sequences aiw~thing might come across. Helps me to parameterize the generic fast line fitting function for each type of .."where am I, what's ahead, what's behind".. decision that the aiw script needs to detect and navigate through. Made about 20 of these tracks so far.... bunch more to do. This is one of those for param writing / testing consecutive same and opposite direction corners close to 180 degrees:
Thanks... my machines at home are allot more fun than those they had me interfaced to at the hospital. Below are images of the first test track I made to test the new vector calculus functions... and the fast line aiw~thing built on it (in ~ 13 sec). This track was mostly to test my new straight-away fitting function, although I threw in a bunch of non-standard corner sequences between the straights. The new function for fitting straights was designed to handle slightly curving straights... which are not as easy to do as regular straight line stuff between same and opposite direction corners. ref: Track is 12m wide... distance between waypoints (little green dots) is about 1 car length (5m). Direction: Clockwise (see xPit / xFinish / xSector arrows). View attachment 18572 View attachment 18573 View attachment 18574 View attachment 18575 View attachment 18576
Yes! DITTO on the Health as that is by far the most important thing to take care of. Look forward to seeing your progress and maybe I can beta test for you this December since we did not get to last December Cheers, Mike
