16 Paper Presentations at ICDCN2026
16 presentations and 2 awards at ICDCN 2026 held in Nara, Japan (January 2026)
Publications & Talks Awards
International Conference · 2026
A Transformer-based Approach for Attribute-Aware Route Optimization Guided by ILP Solutions
Abstract
This paper formulates the Nursing Care Taxi Dispatch (NCTD) and proposes a novel approach that combines Transformer with a heuristic algorithm. Conventional methods struggle with NCTD due to its complex constraints and the frequent risk of violation. To improve the quality, we first obtain relatively high-quality solutions based on Integer Linear Programming (ILP), and use them as training data to learn a Transformer-based architecture. By incorporating user attributes strongly related to constraint satisfaction, the model can well capture the path features of ILP solutions without constraint violation. It also uses the masking mechanism and a heuristic algorithm to address all constraints. We evaluate our approach on datasets created from real-world operational data in two Japanese cities and compare it with existing machine learning-based methods in terms of quality, execution speed, and violation rate. Results show that our method can reduce the mean travel time by 1-2 minutes for problems with fewer than 30 nodes, and provides a good balance between quality, execution time, and violation rate.
Related Lab Activities
Related Projects
A Platform for Digitalizing Knowledge of Regional Communities
Japan Science and Technology Agency (JST) CREST
「基礎理論とシステム基盤技術の融合によるSociety 5.0のための基盤ソフトウェアの創出」領域
