Monday, August 17, 2015

Back to the Future

Ok so this feels vaguely familiar all over again. I am starting classes as of today. It feels strange and exciting at the same time.

I am currently enrolled in the Artificial Intelligence for Robotics. This class is also known as Programming a robotic car. The more I'm taking the lessons the more excited I am getting about the course.

I went through the first lesson and problem set, we learned how a localization function works. I got a refresher on Bayes rule. I am sure I have a lot of learning to go, but so far I'm really optimistic.

One thing that was tripping me up a lot was figuring out which points were which in my inexact motion calculation. Additionally I was needed to make sure i accounted for Total Probability. In this case I am not going to normalize, I just multiply by all the probabilities at each of the squares as necessary.

Inexact motion is basically a way to assign probabilities to where on a grid you have moved to based on your starting point, and direction. If you overshoot (meaning you should have stopped at grid spot X but you went to gridspot X+1 assuming rightward movement), it's one thing, undershoot is less, and exact is right on.

In the problem set we have a similar inexact motion condition, except in this case it's undershoot or right on. Additionally what i was doing before was moving everything THEN doing the calculation, this was throwing things off.

All in all it's a fun first lesson. Also I have heard there will be an optional hardware component....I am SOOOO excited for that, I love that part of software, when you get to interact with things.