Bai Liu (刘柏)

bailiu [at]   [CV]   [LinkedIn]   [GitHub]  

I am currently a Ph.D. candidate in Laboratory for Information and Decision Systems (LIDS), Massachusetts Institute of Technology, advised by Prof. Eytan Modiano.

My research focuses on designing practical algorithms to solve the learning and control problems in network systems, with the application of optimization, machine learning and inference methods.



  • Bai Liu, and Eytan Modiano. Tracking MaxWeight*: Optimal Control for Partially Observable and Controllable Networks. IEEE/ACM Transactions on Networking. Submitted.
  • Bai Liu, Qiaomin Xie, and Eytan Modiano. RL-QN: A Reinforcement Learning Framework for Optimal Control of Queueing Systems. ACM Transactions on Modeling and Performance Evaluation of Computing Systems. Submitted.   [arXiv]
  • Bai Liu, and Eytan Modiano. Optimal Control for Networks with Unobservable Malicious Nodes. Performance Evaluation, to appear on November 2021
  • Bai Liu, Qiaomin Xie, and Eytan Modiano. Reinforcement Learning for Optimal Control of Queueing Systems. 57th Annual Allerton Conference on Communication, Control, and Computing, September 2019.
  • Bai Liu, Ke Han, and Jianming Hu. Global Optimization Framework for Real-time Route Guidance via Variable Message Sign.   [arXiv]


  • Jianming Hu, Xin Pei, Bai Liu, et al. An Information Distribution Method of Variable Message Sign Based on Prediction Method. Chinese Invention Patent. Publication Number: CN105303856A. Publication Date: 2016.02.03.

Software Copyright

  • Intelligent Networking Transportation Guidance System Platform [INGSP] V1.0. Computer Software Copyright. Registration Number: 2016SR252223. Date: 2016.06.01.

Selected Research Projects

Optimal Control for Networks with Unobservable Malicious Nodes

Modern network systems often offer limited access and suffer from adversarial attacks. In this project, we focus on networks with unobservable malicious nodes, where the network dynamics, such as external arrivals and control actions of malicious nodes can be adversarial.

Optimal Control for Partially Observable and Controllable Networks

Modern networks are complex and may include components that cannot be fully controlled or observed. We propose a control algorithm for such overlay-underlay networks where the network controller can only observe and operate on overlay nodes, and the underlay nodes are neither observable nor controllable.

Reinforcement Learning in Network Control   [arXiv]

By applying model-based reinforcement learning and Lyapunov analysis, we design an algorithm for queueing networks with unbounded state spaces. We prove that under our algorithm, the average queue backlog can get arbitrarily close to the optimal result.

Dynamic Transportation Network Modeling   [arXiv]

Aiming at improving VMS (variable message sign) display strategy, we formulate transportation network model with feedback scheme. We then design optimization algorithm with linear decision rule and heuristic optimization approach. A simulation case study is conducted on a real-world test network in China, which shows the advantage of the proposed adaptive VMS display strategy.