PGM for Vehicle Route Optimization with Reinforcement Learning
A project at the intersection of probabilistic graphical models and RL for vehicle routing.
This project explores how probabilistic graphical models (PGMs) and reinforcement learning (RL) can be combined for vehicle route optimization. The codebase includes baseline and advanced RL implementations, route visualizations, and benchmarking artifacts.
- What I built: end-to-end experiments for route planning policies, including environment design and training/evaluation loops
- Core idea: use probabilistic structure and RL objectives to improve route quality under operational constraints
- Outputs: reproducible experiments, training diagnostics, and route/vehicle-load plots
Poster
Links
- GitHub: AKobeissi/pgm-reinforcement-learning
- Reference paper: Towards Reinforcement Learning over State and Temporal Abstractions for Vehicle Routing Applications
- Poster (PDF): PGM vehicle routing poster
- Poster/script: benchmark/poster.py