These codes are part of course project completed with zhengjilai and Lsc-cs aiming to provide a Ride-Hailing Platform with simulator. Concretely, for orders dispatching, we constructs a bipartite graph for orders and drivers with weights of edges calculated by a value map processed from historical data using dynamic programming (inspired by Large-Scale Order Dispatch in On-Demand Ride-Hailing Platforms: A Learning and Planning Approach), and solves the matching problem by KM algorithm. Fleet management algorithm is also based on the value map.
The simulator codes are modified from here. We use orders data from DiDi (Nov 2016, Xi’an City Second Ring Road Regional Trajectory Data Set).
- Python 3
- Process data.
python data_processing/make_data.py - Run simulator.
python run/run_fix_drivers_number_with_value_map_with_fleetmanagement.py