Dependencies: TensorFlow, OpenAI Gym, Roboschool v1.1
The only file that you need to look at is run_expert.py, which is code to load up an expert policy, run a specified number of roll-outs, and save out data.
In experts/, the provided expert policies are:
- RoboschoolAnt-v1.py
- RoboschoolHalfCheetah-v1.py
- RoboschoolHopper-v1.py
- RoboschoolHumanoid-v1.py
- RoboschoolReacher-v1.py
- RoboschoolWalker2d-v1.py
Instructions: http://rll.berkeley.edu/deeprlcourse/f17docs/hw1fall2017.pdf
Plotting handout: http://rll.berkeley.edu/deeprlcourse/docs/plotting_handout.pdf