- Open this Google Colab and make a copy.
- Running the first 4 cells will take care of all dependencies you will require.
- Look for the spots that say "YOUR CODE STARTS HERE". There should be 3 functions you are required to fill in.
- Open this Google Colab and make a copy.
- Look for the spots that say "Code starts here". There should be 6 functions you are required to fill in.
- cd into csc375-w22-p3 in your cloned github directory.
- Run
pip install pybullet
in your command line shell. - Look for "YOUR CODE STARTS HERE". There should be The only two files containing such statements are in
main.py
androbot.py
and 6 places in total. - You should not have to make any changes in any of the other files.
- After you are finished, running
python main.py
should bring up a simulated robot that picks up blocks, places them in the bins, then takes them out of the bin and repeats indefinitely.
- Open P1_differential_flatness.py in HW2.
- Look for the spots that say "Code starts here". There should be 3 spots you are required to fill in.
- Open this Google Colab and make a copy.
- Look for the spots that say "YOUR CODE STARTS HERE". There should be 9 spots you are required to fill in.
- From your command line: cd kuka_grasp_rl.
- Look for the spots that say "Code starts here". There should be 5 spots you are required to fill in in agents.py and kuka_grasp_actor_critic.py.
- Run by calling: python kuka_grasp_actor_critic.py.
- renders=True in the env initialization will enable the GUI when running on your local machine. This will help with debugging.
- After you have debugged, use UTM machines for training in full, by disabling the GUI (renders=False in the env initialization). This will train faster.