Next, we will clarify the theoretical understandings we have gathered so far by observing how environments can be solved by artificial agents. We will see how this can be achieved even through randomly sampling actions from an agent's action space (possible actions an agent may perform). This will help us to understand the complexities involved in solving even the simplest of environments, and why we might want to call upon deep reinforcement learning shortly to help us to achieve our goals. The goal we are about to address is creating a self-driving taxi cab in a reduced, simulated environment. While the environment we will deal with is much simpler than the real world, this simulation will serve as an excellent stepping stone into the design architecture of reinforcement learning systems.

To do this, we will be using OpenAI's gym, an adequately...