Lab: Challenge

Challenge

For this task, try to modify the models given for the one or more of the following robots

Cheetah
Ant
Humanoid
Hopper

For the modification, this can be as simple as playing around with the weights for the neural network, implementing a reinforcement learning algorithm already available in OpenAI Gym to something as complicated as creating your own reinforcement learning algorithm. The choice is yours to make! The code for these agents can be found in the roboschool/agent_zoo directory. The functions all vary a bit, but try to dig through the code and make sure to ask questions. This is a great way to learn how to read other folks' code while learning how to use an open source project.

Questions

What type of algorithm did you use/create? Did you modify the existing algorithm or make your own?
Did you use a reinforcement learning algorithm or did you try another method?
What are some of the pros and cons of the approach you tried?
Would you want to deploy this on a real robot? Why or why not?

Don't forget to discuss these questions and more in our #bio_robots slack channel!