Reinforcement learning is a machine learning paradigm that focuses on developing agents’ abilities to make decisions in dynamic and interactive environments. In RL, an agent learns by exploring the environment, taking action and observing the consequences of its choices. This process is motivated by the concept of rewards and punishments that the agent receives in response to its actions/actions. After presenting the latest versions of the RL algorithm, the article included, depending on the testing environment conditions tested, graphs related to the parameters for evaluating the learning process of an agent using the PPO variant: path length, loss function, cumulative reward value and entropy value. The purpose of the study was to show the latest version of the RL algorithm, which can not only be used in a dynamic change environment, but also in the control of autonomous vehicles.