3 Design custom environments for reinforcement learning algorithms
This chapter covers
- Essential Concepts for designing custom reinforcement learning environments
- Conceptual framework for designing any business optimization environment
- Design a 2D environment for robot navigation in warehouse
- Design a 2D environment for dynamic pricing for perishable products
- Design a 3D environment for trailer/container loading and packing optimization
If you can’t describe what you are doing as a process, you don’t know what you are doing.
W. Edwards Deming, Quality & Process Improvement Guru
Reinforcement learning agents learn from experiences, and trial and error. Training them is like raising a child—they explore, make mistakes, get feedback, and slowly shape better behaviors. But just like children, they often produce a lot of mess in the process. Before learning to walk, a child stumbles countless times. Similarly, a reinforcement learning agent can take thousands—even millions—of wrong actions before discovering an effective policy.