AI Project 4 – GridWorld Exploration using Reinforcement Learning

Objective

In this project P4, you are essentially exploring a gridworld.  (Go North, Go East, Go South, Go West.  Go wherever you want.  Go crazy.)  Just try to maximize your rewards.

T & R

Yeah, those would be nice to know, but unfortunately, we don’t know the T&R values.  We will have to learn those along the way.

How many grid worlds?

There are many environments.  Each environment can be explored up to m times.

What’s the score?

Your score is always being averaged.

What to Submit

Nothing.  Present in class.

Your score is updated via the API.

Use of API

We will play and record the games interactively with each other.  Details of the API will be shared via Slack and discussed in class.

Grading Rubrik

  • 5 points:  Based on your in class presentation.
  • 5 points: Based on your average score.