Hybrid Search with Graph Neural Networks for Constraint-Based Navigation Planning

Marc-Emmanuel Coupvent des Graviers, Kevin Osanlou

Abstract: Route planning for autonomous vehicles is a challeng- ing task, especially in dense road networks with mul- tiple delivery points. Additional external constraints can quickly add overhead to this already-difficult prob- lem that often requires prompt, on-the-fly decisions. This work introduces a hybrid method combining ma- chine learning and Constraint Programming (CP) to improve search performance. A new message passing- based graph neural network tailored to constraint solv- ing and global search is defined. Once trained, a single neural network inference is enough to guide CP search while ensuring solution optimality. Large-scale experi- ments using real road networks from cities worldwide are presented. The hybrid method is effective in solving complex routing problems, addressing larger problems than those used for model training.