AdvancedMazeExercise

Overview

In this exercise your goal is to implement a neural network or genetic algorithm based learner for the maze environment. Since there are two options for this, the goals vary somewhat:

For a Neural Network, you want to create a neural network that will eventually, and reliably reach the goal.

For a Genetic Algorithm, you want a genetic system that will with a reasonable amount runtime, learn a sequence of actions that will solve the maze.

Instructions -- Neural Network

For this one there is almost limitless ways to implement. Some of the simpler solutions involve having the agent simply follow the left hand wall. Keep in mind that you'll want to max out the lifetime so that it will reach the end if it ever will.

Instructions -- Genetic Algorithm

Once again there are a lot of ways to implement a genetic algorithm learner for a maze. One obvious solution you may want to try is having the genetic algorithm ignore all incoming input from the sensors, and simply learn a sequence of actions.