AlphaTransit Uses MCTS and Neural Policy-Value Network for City-Scale Bus Routing

AlphaTransit couples Monte Carlo Tree Search with a neural policy-value network to handle delayed feedback in transit route network design, where local route extensions can create downstream transfer bottlenecks invisible until the full network is assembled. The system targets city-scale bus network optimization.