TensorFlow 1.x Deep Learning Cookbook

How it works...

The preceding code implements a random agent; the agent randomly selects one of the four actions:

Another important thing to observe is that in this environment, the action space is Discrete, while the observation space is of type Box. The terms Discrete and Box in reference to space (action/observation) in OpenAI refer to the allowed values. The Discrete space allows a fixed range of non-negative numbers, in our case, (0,1,2,3). The Box space, on the other hand, represents an n-dimensional box, thus any valid observation would be an array of 210 × 160 × 3 numbers for Pac-Man.