Combining Constructive and Perturbative Deep Learning Algorithms for the Capacitated Vehicle Routing Problem
This work addresses the NP-hard Capacitated Vehicle Routing Problem for logistics and delivery optimization, presenting an incremental improvement by combining existing deep learning approaches with a memory-efficient enhancement.
The paper tackles the Capacitated Vehicle Routing Problem by developing the Combined Deep Constructor and Perturbator, which integrates constructive and perturbative deep learning heuristics with attention mechanisms, and proposes a memory-efficient algorithm that reduces memory complexity by a factor of the number of nodes. It shows cost improvements on common datasets compared to other deep learning methods and achieves results close to state-of-the-art operations research heuristics, while enabling use on instances with over 100 nodes.
The Capacitated Vehicle Routing Problem is a well-known NP-hard problem that poses the challenge of finding the optimal route of a vehicle delivering products to multiple locations. Recently, new efforts have emerged to create constructive and perturbative heuristics to tackle this problem using Deep Learning. In this paper, we join these efforts to develop the Combined Deep Constructor and Perturbator, which combines two powerful constructive and perturbative Deep Learning-based heuristics, using attention mechanisms at their core. Furthermore, we improve the Attention Model-Dynamic for the Capacitated Vehicle Routing Problem by proposing a memory-efficient algorithm that reduces its memory complexity by a factor of the number of nodes. Our method shows promising results. It demonstrates a cost improvement in common datasets when compared against other multiple Deep Learning methods. It also obtains close results to the state-of-the art heuristics from the Operations Research field. Additionally, the proposed memory efficient algorithm for the Attention Model-Dynamic model enables its use in problem instances with more than 100 nodes.