October 2019
Intermediate to advanced
340 pages
8h 39m
English
In Step 3, the state (position and velocity) keeps changing accordingly and the reward is -1 for each step.
You will also see in the video that the car is repeatedly moving to the right and back to the left, but doesn't reach the top in the end:

As you can imagine, the Mountain Car problem is not as easy as you thought. We need to drive the car back and forth to build up momentum. And the state variables are continuous, which means that a table lookup/update method (such as the TD method) will not work. In the next recipe, we will solve the Mountain Car problem with FA methods.
Read now
Unlock full access