I have a couple of suggestions/requests for next season if you think they would be beneficial and/or possible.
1. It's hard to test avoidance code during testing as there is nothing to avoid on the track -
except the walls :) It would be great if we had another monsterborg on track at the same time to avoid. Perhaps a powered off monsterborg in one of the outer lanes (lane +1 or +2) of the start finish straight. That way people wanting to test for speed wouldn't get stuck by sticking to the inside and people wanting to test avoidance would go wide around the final turn and try to avoid a stationary car. The part we all need to work on!
2. Reviewing the logs whilst rewatching the race can be difficult to match up. If your watching the race and see your monsterborg did something strange you weren't expecting, trying to find the area in the log file is difficult. Would it be possible to have the local time down to the second displayed in the lower corner like you have for the split times in testing?
Thank you for everything that you're doing with FormulaPi especially running the races in the hot weather. It would have been very easy to postpone racing until a cooler day or a different time when it wasn't so hot. But not wanting to disappoint a bunch of strangers on the internet you fired up the live stream and pushed on. I think I speak for all the competitors when I say 'thank you, you're awesome!'
Jamie
IT Guy
Couldn't agree more Jamie - I definitely think there is more to be had from testing and have pondered the exact same ideas you are putting forward...I second them!
I second the third point too...if that makes sense ;o)
1) For point one, one idea would be to have 2 race cars, with possibly different codes. You can prepare one for a car to move to some lane and then stop and the other to avoid it. Or have two codes just the same and let them run if you prefer to.
Other idea would be if you can ask for a stopped robot to be placed somewhere in the track, like "put it in the finish line, on lane -2", or "stuck in the first right turn after the left / S turn".
2) For matching times and the streaming we use the stored images. The image shows where the robot was and is logged too. Having the race time on screen and the time since the green light was detected on the logs could help too.
3) I "third" the third point! :D Thanks!
4) I'll add one more point: With submission deadline on Friday and races on Monday/Wednesday there's too little time to work with the information from the latest run, specially for those who have more time on weekends for this. What if submission deadline was on Monday and races on Tuesday / Thursday?
Thanks!
Jorge
The simulator does allow some form of testing of avoidance code. A real word test is a good idea.
We have thought about a round where we place 5 robots randomly on the track. The idea being if you can navigate a lap without collision with any of the 5 robots, you would get 5 points. Of course, if we have a round such as this, we would also need a testing session for it. I'm hopeful all this is going to happen next season if we can fit it in with the list of bits we are doing (there are possibly some really big and awesome changes which I can't wait to scream about).
We are well aware of the horrible Friday submission problem. We'd love to let people work on their code over the weekend and submit code on Monday. The problem is that Claire can't do Tuesdays, so we would end up without her amazing commentary, or worse - with mine :-O
It takes a bit of time to program 20+ 4GB cards, and we like to have them ready to go for calibration.
Ideally we'd also like to have a day between races to fix things just in case something goes wrong (think track, cameras, internet connection, broadcast software - there's quite a bit of stuff behind the scenes).
So, the only other possibility would be Wednesday and Friday (with Thursday as a buffer to accomodate any additional competitors for next season).
I'm not sure about the time stamp. Arron could probably wade in here, but the obvious problem is that the Pi's then need some form of Internet connection to grab the time which opens a whole can of security and cheating worms! Jorge has pointed out - if you have a non accurate time stamp and you log that time when you see the start lights, you should be able to align this to the footage.
Tim
I'm excited for these mystery changes for next season. I've not signed up yet, but I definitely plan to. I've been able to watch some of the races at work and I think I've got a few co-workers interested for next season!
An avoidance round would be great. I really like Jorge's idea about asking to place robots in particular places. That might be difficult if we all requested them in different places though, so a statically placed one might be simpler.
See you all for testing on Monday! Sub 14.80 lap is my goal. Watch out Team Enigma!
I'm really excited if the new faster bots do end up becoming the top class, going to be even more chaos than the MonsterBorgs I think! I'll probably enter both series as I have unfinished business with the monster class(!) but don't want to miss out on the faster class either.
I do have a little concern however that hardware is changing a lot each season which may be hindering the progression of the code. I have big plans for the monsterborg next season given the time to implement and test the code and the addition of the IMU sensors should only add to the capability but the internal GPS positioning could be giving away too much info regarding the bot location(?).
Not sure what other peoples opinions are but I'd like a hardware freeze for a couple of seasons after the next one (I like the idea of a few additional sensors in addition to the camera and the new class looks challenging) but I really want to get heavily involved in the optimisation of the code - the whole reason for releasing the winners code was to have a better starting point each season to build upon. Also the additonal time to play with a standard hardware set would allow people to make bigger changes between seasons that wouldn't be possible mid season (although awesome job Lambda Pi Racing for rewriting the whole thing in Java, brave move!)
Just my two cents.
Adam
I think the lane in the eliminations should be random.
About the Java port and the difficulties I will write something the week.
Add new comment