A new method called CL-GPGLS has been proposed to improve the learning process for large-scale Vehicle Routing Problems.
CL-GPGLS integrates Curriculum Learning (CL) into the Genetic Programming Guided Local Search (GPGLS) method.
This approach uses a predefined curriculum to introduce training instances progressively, starting with simpler tasks and gradually increasing complexity.
Extensive experiments have shown that CL-GPGLS outperforms three baseline methods in solving large-scale VRP effectively.