How an Australian team won the robot soccer world championships | Hacker News Reader

How an Australian team won the robot soccer world championships

smh.com.au

44 pointsbootload8 comments

UNSW has always fielded a strong robotics team, ever since the start of the robocup competition: http://www.cse.unsw.edu.au/about-us/help-resources/for-stude...

They were either first or second the first 5 years of the competition and have generally stayed at the top of the rankings every year.

What's fascinating about Robocup is that each team releases the source code to their robots at the end of each competition, which means each team essentially starts at an equal position every year. What differentiates top performers is their ability to innovate and improve every year.

If the robots are all the same, why are the aussie's robots so much faster? Do they take more of a risk on balance?

The code is not the same; behaviours and inverse kinematics are pretty much never the defaults (they suck). So each team tries to work on things its members are good at, and it seems Australians are particularly good at that.

Ask an ex-member of the Dutch Nao Team anything =)

What sensors are on these things? Is there an API for the whole robot?

Not much; a couple of cameras (which are not in the eyes), a motion sensor, microphone/receiver for audio and a couple of bump sensors.

Aldebaran (the company that makes them) provides the robot with a default OS. The main problem is that it's not really efficient nor stable, so in general teams have to reimplement nearly everything. But to play around with it it's probably decently accessible - as long as you don't do anything really innovative. Aldebaran's website gives more info about it.

What's the current state of team strategies implemented in this competition? Every robot going for the goal on its own? Simple passing forwards? Blocking? Double team on the ball carrier? What's the most number of intended passes in sequence that a team carries out in practice?

Thanks.

In general, afaik, team strategies are not really developed in depth. This is simply because actually governing the robots is exponentially harder than anything else, and a robot that does not know where it is on the field, where the ball is or is unable to move is also useless - no matter what advanced tactic you decide to deploy. And these are REALLY hard problems!

Other problems are: understand which side of the field is actually yours (it's not uncommon to do autogoals, since the field is visually symmetrical), how to understand where you are after a referee repositions you, how to distinguish friendly robots from adversaries, not fall down every five seconds.

And still the game is driven by simple hardware physics: the robots engines tend to overheat very often, which basically results in the robots "melting" on themselves since they can't support their weight for very long. And there how much money a team has to spend comes into play heavily..

So to get back at your questions, not much. In general if you can shoot in the general direction of the enemy's goal more often than your adversaries, you will probably win =)

Reply on news.ycombinator.com