Application Procedures and Challenges of Reinforcement Learning Using Discrete System Simulation

Recently, more and more people have been using reinforcement learning (RL) to solve industrial issues. RL requires long-term trial and error. So, when dealing with large-scaled or complex systems where it’s hard to apply RL directly, discrete system simulation is used as an instead of the real system. This report explained RL using simulation. First, we used a simple logistics model and described the linking simulation and RL. Then, we looked at a more complex industrial case, where we explained the application of RL in the pit crane operation at a waste incineration facility. This study explained the linkage process of simulation and RL in more detail and identified the issues.