Publications

You can also find my articles on my Google Scholar profile.

Conference Papers

Bandit-Guided Dynamic Programming for Last-Mile Delivery in Stochastic Networks

Published in July 29, 2026

We propose a Bandit-enhanced Dynamic Programming algorithm for efficient routing of Autonomous Delivery Vehicles in stochastic urban networks, achieving near-optimal delivery costs with significantly higher computational efficiency than standard Value Iteration.

Recommended citation: Li, Y., & Xiong, X. (2026). "Bandit-Guided Dynamic Programming for Last-Mile Delivery in Stochastic Networks." 2026 INFORMS Transportation Science and Logistics Conference. Cambridge, MA.

Multi-Armed Bandit for Stochastic Shortest Path in Mixed Autonomy

Published in November 22, 2025

We address the exploration-exploitation challenge in autonomous vehicle routing under mixed-autonomy traffic by integrating Upper Confidence Bound strategies into Real-Time Dynamic Programming, achieving provably convergent and computationally efficient routing in large-scale stochastic networks.

Recommended citation: Bai, Y., Li, Y., & Xiong, X. (2025). "Multi-Armed Bandit for Stochastic Shortest Path in Mixed Autonomy." 2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC). 3604-3609.
Download Paper

Intersection Congestion Identification Method Based on Fuzzy C-Means Clustering

Published in July 22, 2025

We propose a Fuzzy C-Means Clustering-based methodology using electronic police data to accurately identify balanced and overall congestion states at urban intersections, validated through a case study demonstrating strong alignment with real-world conditions.

Recommended citation: Li, Y., Bai, Y., & Xiao, B. (2025). "Intersection Congestion Identification Method Based on Fuzzy C-Means Clustering." CICTP 2025: Transportation, Artificial Intelligence, and Energy. 2073-2084.
Download Paper

Intelligent Pixel-Level Segmentation of Pavement Sealed Cracks Using CycleGAN-Based Domain Adaptation

Published in January 9, 2025

We propose a CycleGAN-based domain adaptation approach for pavement sealed crack segmentation, achieving a 6.18% improvement in IoU coefficient over source domain images and enabling effective cross-region pavement distress detection without re-annotation or retraining.

Recommended citation: Li, Y., Chen, J., Lang, H., & Qian, J. (2025). "Intelligent Pixel-Level Segmentation of Pavement Sealed Cracks Using CycleGAN-Based Domain Adaptation." Transportation Research Board 104th Annual Meeting. Washington, D.C.