ROB537 Homework 4: Learning-Based Control


Category: You will Instantly receive a download link for .zip solution file upon Payment


5/5 - (1 vote)

In this assignment, you will apply the principles you have learned thus far to create two learning-based controllers.

Install [Gymnasium]( and familiarize yourself with the gym interface. This should look similar to the interface you used in the last assignment.

1. Create a Q-learning agent that learns to solve the “Cart Pole” environment. The agent should balance the pole for 100 time steps.

– How will you handle the continuous state space?

2. Evolve a neural network to solve the previous task.

– What will you use for your evaluation function?
– What mapping should the network learn?

Your report should include:

– Answers to previous questions
– Performance over time curves for each experiment.
– A description of your algorithms.
– Similarities and differences in performance for the two algorithms + an explanation.

Extra-Credit Bonus: Use these two learning methods to solve another gym task. Choose your favorite environment from either the “Classic Control” or “Box2D” sets.

– What environment did you choose?
– What changes did you have to make to adapt your learning algorithms to this new environment?
– Describe the resulting behaviors each controller learned.

Or Implement a new learning algorithm of your choice like Deep Q-Learning Network. You may choose to implement a new algorithm from the literature or a new algorithm of your own design. Describe your algorithm and compare its performance to the two algorithms above.