Abstract: A path is defined to be followed by a vehicle in an environment represented by a graph of nodes connected by edges. Each node represents the vehicle's position. Each edge between two nodes indicates a transition cost between the two nodes. An instance of the graph, which is represented by a start node, a destination node and mandatory nodes through which the vehicle must pass, is obtained. An initial order of travel through the mandatory nodes is obtained by applying a procedure, using a graph convolutional neural network, adapted to the graph, to the instance. A procedure of local optimisation of the mandatory nodes is executed in order to obtain an optimised order. A path for the instance is defined to be resolved: from the optimised order and for each pair of nodes of the instance to be resolved, from a shortest path between the nodes of the pair.

Abstract: A path is defined to be followed by a vehicle in an environment represented by a graph of nodes connected by edges. Each node represents the vehicle's position. Each edge between two nodes indicates a transition cost between the two nodes. An instance of the graph, which is represented by a start node, a destination node and mandatory nodes through which the vehicle must pass, is obtained. An initial order of travel through the mandatory nodes is obtained by applying a procedure, using a graph convolutional neural network, adapted to the graph, to the instance. A procedure of local optimisation of the mandatory nodes is executed in order to obtain an optimised order. A path for the instance is defined to be resolved: from the optimised order and for each pair of nodes of the instance to be resolved, from a shortest path between the nodes of the pair.