This meeting took place on 6th of February, 2012.

AI

The robot can now chance the ball and go towards the goal once it has it. This is all done in the simulator as the Vision System needs some tweaking before we can use it. The robot can also dribble the ball.

Machine Learning

Martin is making a neural network based Machine Learning algorithm. Finding best way to approach the ball when it’s next to the wall for example. This is a new part of our project so we still need to conduct more research. We were also thinking of possibly having a mixed AI that would be composed of our ”normal” AI and the trained through ML one.

Vision

Now working at 10 to 15 frames per second. It can’t find the direction but Gediminas said he’d try to get that done by Wednesday. Once this is sorted we can start testing the robot’s AI on the pitch.

Robot

Robot base now has omni wheels. Working on getting the kicker more stable. Possibly add sensors on the side such as touch sensors.

Test Bench

We discussed making a benchmark tool for our visual system. This would consist in annotating pictures of the pitch where we have the ball and the robots and comparing these annotations with the ones our vision system gives. It’s a great way of having measures of our system and as we said in the meeting: Philip Wadler loves it.

XP

We decided to split the group in teams of two, each one working on a part of the project. These teams are:

• Laura and Martin on Machine Learning.
• Andrei and Aaron on the Visual test bench.
• Gediminas on Vision.
• Jibran and Michael G. on the Robot (to be changed when they are done and the robot has been revalidated).
• Michael J. and Paul on AI.

It’s a great way of dividing the work and also making sure everyone has something to do.