Most people by now have probably heard of the terms neural network, along with AI, machine learning, and deep learning. But reinforcement learning (RL) is still not quite a household name among the general public. Majority of machine learning advances thus far have involved what is referred to as supervised learning, which involves the use of labeled training data and wherein the outputs generated are based on the mapping of example input-output pairs. But reinforcement learning is now becoming more than just an academic research topic.
Presently, there are relatively few patents and patent applications (P/PA) that are directed to RL systems and processes. We discuss below four examples of RL-related patents. We chose those four RL-related patents primarily because they either cover fundamental aspects of RL, or they are directed to the use of RL in very important types of real-world applications.