Implement the Q-Learning algorithm to the agent

a-maa / RL15

Godspeed everyone

0 stars 1 forks source link

Implement the Q-Learning algorithm to the agent #3

Open MinimalistSwan opened 3 days ago

MinimalistSwan commented 3 days ago

The following parameters will be considered for the model:

learning rate
exploration rate
discount rate

make sure to explain in the comments what every parameters means and how they affect the agent's ability to learn.

MinimalistSwan commented 3 days ago

We should also consider how many times we want to train the agent and how we are going to document and compare the parameters against different values.

we also of course have to decide:

goal reward
penalty for hitting a wall
penalty for every taken step
reward for reaching the sub goal

MinimalistSwan commented 14 hours ago

most of the code the the q-learning algorithm should be there. However, it's abundantly clear that the sub goal is still not implemented properly. Currently, it's a necessity to pass through it. But I think this should be changed to the agent wanting to go there itself in order to get a higher reward. WILL LOOK INTO THIS.