The simulation design layer turns a business process model into a functioning reinforcement-learning environment. Its design choices determine what the agent observes, what it can do, how it is rewarded, and how the system responds. section 3.1
Translate the operational problem into:
- States that capture decision-relevant information, such as a delivery driver’s location, remaining stops, shift time, and traffic. section 3.1
- Actions that fit real-world constraints, such as choosing the next delivery or deciding how many units to restock. section 3.1
- Rewards based on business KPIs, plus transition behavior that specifies how the system responds to actions. section 3.1
