| CPC G01C 21/3438 (2013.01) [G01C 21/3492 (2013.01); G06N 3/08 (2013.01); G06Q 10/02 (2013.01); G06Q 50/40 (2024.01)] | 20 Claims |
|
1. A method comprising:
receiving map data associated with an interested area, travel request data associated with one or more travel requests, and service vehicle data associated with one or more service vehicles;
determining one or more constraints based on the travel request data;
abstracting the travel request data, the service vehicle data, and the map data into a request-vehicle graph comprising nodes and edges, the nodes indicating the service vehicles, onboard items, and the travel requests, and the edges comprising request-vehicle edges, onboard-item-vehicle edges, and request-request edges;
trimming the request-vehicle graph into partial request-vehicle graphs for each service vehicle, wherein the partial request-vehicle graphs comprise a subset of the nodes and the edges of the request-vehicle graphs;
encoding the partial request-vehicle graphs through a graph neural network;
training the graph neural network to predict actions for each service vehicle, by maximizing an amount of requests served and minimizing total matching costs of request assignments to the service vehicles according to the map data, where each travel request and each service vehicle satisfy the constraints; and
operating the service vehicles using instructions based on the actions.
|