publications

Publications in reverse chronological order. Click a thumbnail or the Poster/Demo/Figure buttons to enlarge; Abs shows the summary.

2026
NeurIPS
Poster

Learning Collision-Free Dispatch Policies for Route-Wise Decision-Dependent Anomaly Detection

Xin Xia, Chen Zhang, and Kaibo Liu
Advances in Neural Information Processing Systems (NeurIPS), 2026
AbsPoster
A dispatch policy for UAV fleets that assigns destinations one vehicle at a time, so joint moves are collision-free by construction and decoded in linear time, with guarantees on coverage, focusing, and detection delay. Validated on the Robotarium and on a wildfire rebuilt from satellite frames.
INFORMS
Demo

SCAN: Networked Multi-Agent Reinforcement Learning for Decentralized Anomaly Detection in Graph-Structured Systems

Xin Xia and Kaibo Liu
INFORMS Annual Meeting, 2026 (journal version in preparation)
AbsDemo
INFORMS QSR Best Refereed Paper Award, 2026
Each drone learns from its own evidence and a few messages from neighbours, moves safely under mobility constraints, and the team raises a calibrated alarm. Demonstrated on wildfire and traffic cases with a drone fleet in a 3D simulator.
Technometrics
Figure

Multi-Agent Reinforcement Learning for Anomaly Detection with Hierarchical Adaptive Sampling Strategy

Xin Xia, Jie Guo, Ziqian Zheng, Yujie Ma, Chen Zhang, and Kaibo Liu
Technometrics (conditionally accepted)
AbsFigure
Finalist, IISE QCRE Best Track Paper Competition, 2025
Limited monitoring resources are allocated across many data streams by a hierarchy of learned agents, so that the streams most likely to carry the change are observed first.
IJDS

KAPPA: A Generic Patent Analysis Framework with Keyphrase-Based Portraits

Xin Xia, Yujin Wang, Jun Zhou, Guisheng Zhong, Linning Cai, Kaibo Liu, and Chen Zhang
INFORMS Journal on Data Science (major revision)
AbsCode
Patents are summarised into keyphrase portraits generated by a pretrained language model, which then support retrieval, clustering, and trend analysis.
2025
IEEE TEM
Poster

A Deep Reinforcement Learning Method Solving Bilevel Optimization for Product Design Considering Remanufacturing

Yujie Ma, Xin Xia, and Chen Zhang
IEEE Transactions on Engineering Management, 72, 573-590, 2025
AbsDOIPoster
Product design as the leader and remanufacturing as the follower in a Stackelberg game, solved by a bilevel SAC algorithm (BSAC) with convergence and optimality guarantees.
IJPE

Bilevel Joint Optimization for Product Design Changes with a Resilient Supply Chain Based on Deep Reinforcement Learning

Yujie Ma, Xin Xia, Peng Liu, and Chen Zhang
International Journal of Production Economics, 290, 109791, 2025
AbsDOI
Design changes and the supply chain that must absorb them are optimised together under disruption risk.
Under review
M&SOM

Strengthening Supply Chain Resilience Through Coordinating Optimization of Product Design and Inventory Management: A Multi-Agent Deep Reinforcement Learning Approach

Yujie Ma, Xin Xia, Chen Zhang, and Nan Chen
Manufacturing & Service Operations Management (submitted)
Abs
A multi-agent deep reinforcement learning approach in which each echelon learns to cooperate with its upstream and downstream neighbours.
IISE Trans.

Quality-Driven Multi-Agent Coordination in Multi-Stage Manufacturing Systems: A Distributionally Robust Bilevel Optimization Framework with Reinforcement Learning

Yujie Ma, Jie Guo, Xin Xia, and Chen Zhang
IISE Transactions (under review)
Abs
A distributionally robust bilevel optimization framework with reinforcement learning for coordinating quality decisions across manufacturing stages.
Preprints
arXiv
Preprint

Multi-Agent Causal Discovery Using Large Language Models

Hao Duong Le, Xin Xia, Haijie Xu, and Chen Zhang
arXiv:2407.15073, 2024
AbsarXiv
Several LLM agents with different roles debate and vote on causal edges, combining metadata with observational data.
Patents