The news is out - last year's spectacular and nerve-racking semi final is bound to get a sequel!

Having fought our way through 5 group rounds, great ups and, well let's call them 'memorable' downs, we made the cut and the opponent du demain (well at least the first one, that is) could hardly be more thrilling and daunting at the same time: the HULKs from Hamburg. So we got our work cut out for us. We collected a whole lot of data with Eindhoven on our mind but that certainly doesn't mean that we're not going to try whatever is possible (and maybe even then some!) to win! If you, dear reader, can't remember the showdown battle royale of Bordeaux, check out the Youtube video I linked on the bottom of this post.

Living la Vida Loca

Well, it's a never ending story. In order to get where you want to go, you need to know were you are. Maybe even where you came from in the first place, but that's bonus information for reasonable people who have their life together and actually know how taxes work. For a robot, there are much more fundamental concerns:

When do I turn my head and how fast?
Where is the ball?
Where am I and if I'm not alone - how many of me are there?
Am I unique if and only if I occupy a unique agent state?
What is love?
Where is the ball?
Baby don't hurt me.
What is that line over there?
Where is the BALL?

We got into the finer details of localization logic. For instance, as I teased in the first blog post, we're now looking for the ball while getting up under external assumptions where the ball is most likely to be found. Robots that get a lot of action als tend to need a lot of time face down on the field - but falling can be messy and the NAO is not famous for its gracefully feline intuition. So what do? Well, in addition to our revised ball search after the fall, we've tweaked and calibrated various parameters when to - and crucially, when not to - look for the ball. A striker usually doesn't need to run all that quickly, as long as he's consistent; but more control could mean losing the ball less since dribbling would be more precise and we could avoid unnecessary ball-searches. A robot can spontaneously lose sight of the ball, and adjusting the window of blind trust can be pretty