Selected Presentations

Recent research themes

Nonconvex policy optimization problems in optimal and robust control often possess hidden convexity. Extended convex lifting provides a common framework for explaining when stationary points are globally optimal.

• Extended Convex Lifting for Policy Optimization in Control

  • Department Colloquium, Department of Mathematics and Statistics, Queen's University, March 27, 2026. (Seminar)

• Other seminar versions: University of Toronto, March 2026, UCSD Optimization for ML and AI Seminar, February 2026, and UCSD MAE Dynamic Systems & Controls Seminar, May 2025.

• Shorter introduction: Benign Nonconvex Landscapes in Optimal and Robust Control, 2024 Information Theory and Applications Workshop.

• Earlier work on LQG policy optimization: NUS seminar, 2023, ACC workshop, 2023, UIUC seminar, 2022, University of Michigan seminar, 2022, and Tsinghua seminar, 2021.

• Guest lecture at Harvard University, March 2022: Sample Complexity of Model-based Linear Quadratic Control: LQR and LQG

Weak convexity connects nonsmooth optimization with practical descent and bundle methods that retain rigorous convergence guarantees.

• A Proximal Descent Method for Nonconvex Policy Optimization

  • Invited talk, 62nd Allerton Conference on Communication, Control, and Computing, September 17, 2026. (Conference talk)

• Error Bounds, PL, and Quadratic Growth for Weakly Convex Functions, and Linear Convergences of Proximal Point Methods

  • 2024 INFORMS Optimization Society Conference, March 24, 2024. (Conference talk)

Chordal sparsity and matrix decomposition make large semidefinite and sum-of-squares optimization problems more tractable.

• Chordal Graphs, Semidefinite Optimization, and Sum-of-squares Matrices

  • MAE Seminar, Arizona State University, October 18, 2023.

• Scalable Semidefinite and Polynomial Optimization via Matrix Decomposition,

  • LANS Seminar Series, Argonne National Laboratory, March 23, 2022.

• Focused conference talks: Sum-of-squares Chordal Decomposition of Polynomial Matrix Inequalities, INFORMS 2023; and Block Factor-width-two Matrices and Their Applications to Semidefinite and Sum-of-squares Optimization, INFORMS 2022.

• Earlier seminars: UCSD, 2021, KTH, 2019, and MIT LIDS, 2018, and University of Birmingham, 2017.

• Integrating Autonomy into Traffic Systems: Scalable Control and Optimization

  • INTR 6800 Seminar, Hong Kong University of Science and Technology, November 8, 2021.

• Smoothing Traffic Flow via Control of Autonomous Vehicles

  • Center of Innovation Automation, Volkswagen, Wolfsburg, December 2019; Harvard University group seminar, October 2019.

• Workshop version: Integrating Autonomy into Traffic Systems: Scalable Control and Optimization, IEEE Intelligent Vehicles Symposium Workshop, 2020.

Selected posters

• Benign Nonconvex Landscapes in Optimal and Robust Control

  • NSF Workshop on Reinforcement Learning, January 2025.

• Convex Approximations for a Bi-level Formulation of Data-Enabled Predictive Control

  • Learning for Dynamics and Control, 2024.

• Error Bounds, PL Condition, and Quadratic Growth for Weakly Convex Functions

  • Learning for Dynamics and Control, 2024.

• Sample Complexity of Linear Quadratic Gaussian Control for Output-Feedback Systems

  • Learning for Dynamics and Control, 2021.

• Large-scale Autonomy: Scalable Sparse Optimization and Distributed Control

  • IEEE Conference on Decision and Control, 2019; Harvard SEAS Postdoc Recognition Event, 2019.