Utilizing a discrete event simulation of material handling plans to calculate reinforcement learning rewards
Date of Award
Master of Science (MS)
Computer Science : Modeling and Simulation
Mikel D. Petty
James J. Swain
Discrete-time systems--Simulation methods., Reinforcement learning., Materials handling.
In this thesis, a discrete event simulation model of Steelcase's Athens facility is constructed to simulate material handling plans generated by an external reinforcement learning algorithm. The results of this simulation are consolidated into key performance parameters and provided as a reward to the reinforcement learning algorithm. The model interfaces developed for the reinforcement leaning algorithm are verified through several sets of tests. Validation of the model is conducted in two separate approaches, input model validation using truth data and output data validation using key performance parameters.
Preston, Ryan A., "Utilizing a discrete event simulation of material handling plans to calculate reinforcement learning rewards" (2017). Theses. 215.