Built a custom RL environment in Python/Gymnasium to benchmark PPO against a Destination Dispatching algorithm. 20 floors, 4 elevators, probabilistic traffic patterns with morning/evening rush simulation.
Key result: 84% reduction in average passenger wait time. Currently adding realistic elevator physics (acceleration curves, door cycles).
Github: https://github.com/jonas-is-coding/elevator-ai
Show HN: PPO agent reduces elevator wait times by 84% vs. classical dispatching | Heykuki News