This repository shows the application of reinforcement learning to solve a particular production logistics problem dealing with the coordination of material supply for an assembly process.
Train an ACER-agent (by default for only 10,000 steps, which should take <1min):
docker-compose up acer-training
Plot performance (might require additional steps for connecting the display):
docker-compose up acer-plot-training
Tilemap-animation of the trained agent (visit http://localhost:5000):
docker-compose up acer-web-animation-tilemap