Deep reinforcement learning (DRL) (Glossary)

Deep reinforcement learning (DRL) belongs to the category of artificial intelligence and is also used in the fields of automation, industry and Factory 4.0. It is a modern method of training machines to solve tasks independently by trial and error and to continuously improve.

In contrast to conventional artificial intelligence, where fixed rules are specified, a system with deep reinforcement learning (DRL) learns on the basis of experience. The machine is given a task and tries out different actions. Correct decisions are rewarded - mistakes are penalised. In this way, the system gradually finds the best way to master a task.

An illustrative example: In a modern factory, a robot arm controls the assembly of car parts. Initially, the robot makes a lot of mistakes. However, with the help of deep reinforcement learning (DRL), it learns with every movement and becomes increasingly precise and faster. In the end, the robot works independently and reliably.

Deep reinforcement learning (DRL) helps companies to automate processes and significantly increase their efficiency.