AlphaTransit: AI Framework for City-Scale Bus Network Design
AlphaTransit is a search-based planning framework introduced for city-scale bus network design, addressing the Transit Route Network Design Problem (TRNDP). The framework uses Monte Carlo Tree Search (MCTS) coupled with a neural policy-value network to guide route construction under delayed simulator feedback. The policy proposes route extensions, while the value estimates downstream design quality, enabling decision-time lookahead without running simulator rollouts inside the search tree. This approach tackles the challenge where route extension quality is only visible after the full network is assembled, and local improvements can create transfer bottlenecks, redundant overlap, or reduce overall throughput. The paper is published on arXiv with ID 2605.28730.
Key facts
- AlphaTransit is a search-based planning framework for city-scale bus network design.
- It addresses the Transit Route Network Design Problem (TRNDP).
- The framework couples Monte Carlo Tree Search (MCTS) with a neural policy-value network.
- The policy proposes route extensions, the value estimates downstream design quality.
- It provides decision-time lookahead during route construction without simulator rollouts.
- Route extension quality is often visible only after the full network is assembled.
- Local improvements can create transfer bottlenecks, redundant overlap, or reduce throughput.
- The paper is published on arXiv with ID 2605.28730.
Entities
Institutions
- arXiv