HW 2 - Q learning vs Sarsa

[nickumia]

Notes:

1. Use the same Gridworld environment. Show the picture of your own Grid world including start/final state, blocks.
   • [x] Complete the program code (refer to hw2-code.py file).
   • [x] Complete GridEnv class
     • [x] Insert all information of your Grid environment in class GridEnv
     • [x] constructor (init), returns start state of your grid.
     • [x] implement step function in class GridEnv. ‘step’ function reads action and returns next-state, reward, done, info
     • [x] implement reset function in class GridEnv. ‘reset’ function moves agent back to start state. ‘reset’ function returns start state.
2. Implement Q learning - You have to repeat the loop part of pseudo code in slide p. 33 using many episodes.
   • [x] for each episode
     • [x] For each step in episode
       • [x] Update Q value using q learning formula in slide p. 33
     • [x] show the results of Q values
     • [x] show the results of policy
3. Implement Sarsa - You have to repeat the loop part of pseudo code in slide p. 22 using many episodes.
   • [x] for each episode
     • [x] For each step in episode
     • [x] Update Q value using Sarsa formula in slide p. 22
   • [x] show the results of Q values
   • [x] show the results of policy

[nickumia]

The code for this works (and, presumably, is complete). However, the logic should be documented because I can't say that I fully understand how I got it to work haha..

[nickumia]

Initial Report is uploaded.. but further documentation will be created in the future. 😮‍💨