Evaluating the performance

The PPO agent is trained by the following command:

python train_test.py

Once the training is complete, we can test the agent by setting the following:

train_test = 1

Then, we will repeat python train_test.py again. On visualizing the agent, we can observe that the car first moves backward to climb the left mountain. Then it goes full throttle and picks up enough momentum to drive past the steep slope of the right mountain with the flag on top. So, the PPO agent has learned to drive out of the mountain valley successfully.

Full throttle

Note that we had to navigate backward first and then step on the throttle in order to have sufficient momentum to escape gravity and successfully drive out of the...