Tuesday, 12. January 2016
Yesterday, on monday the 11. january, we had the big final at the HSR. Unfortunately we didn't win, but in comparison to other solutions, our algorithm wasn't bad. Without a little mistake (we know now), it might have been the best ;-)
But first, congratulations to the group 'FasterThanLight' for winning the final. Well done!
As already mentioned the final didn't went very well. But not only for us. All groups had there problems... We all had too little real life experience. If you just test your algorithm against a simulator over months, you will probably have some little problems on a real track. There are important things you have to know, we couldn't predict by working with the simulator.
The little problems
First we had little problems with the start-speed, which was too high. Funnily enough, almost all groups had that problem first :-) With the simulator you never have these problems... However, if the initial speed is too high on the real track, it ends with behavior like this:
But after a little correction of our starting algorithm, that problem was solved.
The big problem(s)
The main problem in the final was our position correction logic. Sometimes it happens that you lose your position on the track because of some irregularities in the GYRO-values. (for example when you drift, etc...) For that case you need some logic to find the position again. For that logic we used data from the light barriers, which worked in the simulator perfectly, but they are not reliable on the real track.
That problem caused that our logic did not found the position again or detected a wrong position. Because of the wrong position the algorithm mixed the straight parts in the speedup phase. If the algorithm thinks he is on the long straight part, but he is on a short one, he speeds up too much. Thats why we flew of the track in the final round.
One promising training
But first we had a training round, which started very promising. The algorithm worked very well until... guess what. We lost our position. See the good start of that round yourself:
So, even we didn't win, we saw that our algorithm wasn't that bad :-) I think with a few more rounds on the track and some small corrections, it would work very good.
After the final we had our last presentation of the project. I think that went very well... You find the slides to our presentation here.
It was a very interesting project and we learned a lot. It included very exciting topics, like: