Able to generate training data + using ReLu instead of linear activation and decreased alpha

Now functional with dynamic subgoals.

Removed Burlap code

Added proper comments to OrthogonalSubGoals and cleaned up code

Fixed out of bounds issues DeepQLearner when using randomly generated map + OrthogonalGoals now compatible with multi agent system. Do note that agents can only cut goal clockwise.

Now able to use full feature vector as input. Todo: update MLP for each subgoal instead of only "WW"

Implemented first stages of Q-Learning

Refractored most of OrthogonalSubGoals. More robust implementation of goal distance to center of fire implemented.

Closest goal to the agent is selected

Orthogonal goals now set by random initialised MLP

Cleaned up code and added comments

altered rounding error cosine + dotproduct function + inputcosyne function

CHANGED features

Changed function winddirection and function distanceFromCenterToFireline + added new function for combined input

HRL fully functional and documented