Guangyi Liu's Home Page

Photo by Christa Neu

Guangyi Liu

Hi, welcome to my website. I'm currently a Postdoctoral Research Scientist at Amazon Robotics. I obtained my Ph.D. at the Autonomous and Intelligent Robotics (AIR) Laboratory, Department of Mechanical Engineering, Lehigh University, working under the advisory of Prof. Nader Motee. Before joining Lehigh, I obtained my B.E. degree at Beijing Institute of Technology. My research interests are risk and robustness analysis in networked control systems and perception systems. I served as a workshop organizer in ACC 2023 and organized four invited sessions at ACC 2023, 2024, and 2025.

Email: gliu@lehigh.edu Google Scholar

Research Areas:

Distributionally Robust Multi-Agent Reinforcement Learning for Dynamic Chute Mapping

(Paper: P5) We propose a Distributionally Robust Multi-Agent Reinforcement Learning (DRMARL) framework for destination-to-chute mapping in Amazon Robotics warehouses, designed to handle uncertain and dynamic package induction rates. DRMARL integrates group distributionally robust optimization (DRO) with a contextual bandit-based predictor to improve learning efficiency and ensure robust chute mapping under varying induction conditions.

Beyond Uncertainty: Risk-Aware Active View Acquisition for Safe Robot Navigation and 3D Scene Understanding with FisherRF

(Paper: C12) We introduce a novel approach to enhance robot navigation safety and improve 3D scene understanding by extending beyond conventional uncertainty-based methods.

Video

Risk analysis of cascading failures in networked control systems with autonomous vehicles and multi-agent rendezvous.

(Papers: J2, C5, C6, C7, C8) We investigate and quantify how the existing failure (e.g., inter-vehicle collisions) will propagate through the network. We also show how communication time-delay and input uncertainty will pose fundamental limits and trade-offs on the risk of cascading collisions.

Data-Driven Distributionally Robust Mitigation of Risk of Cascading Collisions

(Papers: C10, C2) We introduce a novel data-driven method to mitigate the risk of cascading failures in delayed discrete-time Linear Time-Invariant (LTI) systems, which involves formulating a distributionally robust finite-horizon optimal control problem while satisfying a set of distributionally chances constraints on cascading failures.

Symbolic Perception Risk in Autonomous Driving

(Papers: C9, C3) We develop a novel framework to assess the risk of misperception in a traffic sign classification task in the presence of exogenous noise.

Risk of Phase Incoherence in Wide Area Control of Synchronous Power Networks

(Papers: J1) We develop a framework to quantify systemic risk measures in a class of Wide-Area-Control (WAC) laws in power networks in the presence of noisy and time-delayed sensory data, and reveal the effect of network parameters, information flow in WAC architecture, statistics of noise, and time-delays are characterized.

Robust Analysis of Multi-agent Map Classification with RNN

(Papers: C4, C1, P1 - P4) For a given stable recurrent neural network (RNN) that is trained to perform a classification task using sequential inputs, we quantify explicit robustness bounds as a function of trainable weight matrices.

News:

(5/3/2025) I'm going to give a talk at 2nd NESCW about the DRMARL work at Amazon Robotics.
(5/1/2025) Our work DRMARL at Amazon Robotic is accepted at ICML 2025: (Preprint: Distributionally Robust Multi-Agent Reinforcement Learning for Dynamic Chute Mapping.)
(3/12/2025) Our recent work at Amazon Robotic is available online: Distributionally Robust Multi-Agent Reinforcement Learning for Dynamic Chute Mapping.
(1/18/2025) Our invited session on "Autonomous Risk-Aware Perception, Planning, and Control", co-organized with Michael M. Zavlanos, Ufuk Topcu, and Nader Motee got accepted at ACC 2025!

Media:

"Innovative algorithms for safer robot perception" interviewed by P.C. Rossin College of Engineering and Applied Science, Lehigh University.

Journal Publications:

[J2]. Guangyi Liu, Christoforos Somarakis, and Nader Motee. "Risk of Cascading Collisions in Network of Vehicles with Delayed Communication", (under review at IEEE-TAC).

[J1]. Christoforos Somarakis, Guangyi Liu, and Nader Motee. "Risk of Phase Incoherence in Wide Area Control of Synchronous Power Networks.", IEEE-TAC.

Conference Proceedings:

[C13]. Guangyi Liu, Suzan Iloglu, Michael Caldara, Joseph W. Durham, Michael M. Zavlanos. "Distributionally Robust Multi-Agent Reinforcement Learning for Dynamic Chute Mapping.'' , ICML 2025.

[C12]. Guangyi Liu*, Wen Jiang*, Boshu Lei*, Vivek Pandey, Kostas Daniilidis and Nader Motee. "Beyond Uncertainty: Risk-Aware Active View Acquisition for Safe Robot Navigation and 3D Scene Understanding with FisherRF'' (* equal contribution) (submitting to RAL)

[C11].Vivek Pandey*, Arash Amini*, Guangyi Liu, Ufuk Topcu, Qiyu Sun, Kostas Daniilidis, Nader Motee. "Scalable Networked Feature Selection with Randomized Algorithm for Robot Navigation.'' (* equal contribution) IROS 2024.

[C10]. Guangyi Liu, Arash Amini, Vivek Pandey, and Nader Motee. "Data-Driven Distributionally Robust Mitigation of Risk of Cascading Failures.", 2024 American Control Conference (ACC).

[C9]. Guangyi Liu, Disha Kamale, Cristian-Ioan Vasile, and Nader Motee. "Symbolic Perception Risk in Autonomous Driving.", 2023 American Control Conference (ACC), May 31 - June 2, San Diego, CA, USA.

[C8]. Guangyi Liu, Vivek Pandey, Christoforos Somarakis, and Nader Motee. "Cascading Waves of Fluctuation in Time-delay Multi-agent Rendezvous", 2023 American Control Conference (ACC), May 31 - June 2, San Diego, CA, USA.

[C7]. Guangyi Liu, Christoforos Somarakis, and Nader Motee. "Emergence of Cascading Risk and Role of Spatial Locations of Collisions in Time-Delayed Platoon of Vehicles.", The 61st IEEE Conference on Decision and Control (CDC), December, 2022, Cancun, Mexico.

[C6]. Guangyi Liu, Vivek Pandey, Christoforos Somarakis, and Nader Motee. "Risk of Cascading Failures in Multi-agent Rendezvous with Communication Time Delay.", 2022 American Control Conference (ACC), June 8-10, Atlanta, GA, USA.

[C5]. Guangyi Liu, Christoforos Somarakis, and Nader Motee. "Risk of Cascading Failures in Time-Delayed Vehicle Platooning.", The 60th IEEE Conference on Decision and Control (CDC), December 13-15, 2021, Austin, Texas, USA.

[C4]. Guangyi Liu, Arash Amini, Martin Takáč, and Nader Motee. "Classification-Aware Path Planning of Network of Robots." THE 15TH INTERNATIONAL SYMPOSIUM ON DISTRIBUTED AUTONOMOUS ROBOTIC SYSTEMS 2021 (DARS 2021), June 2021, Kyoto, Japan.

[C3]. Arash Amini, Guangyi Liu, Vivek Pandey, and Nader Motee. "Impact of Misperception on Emergence of Risk in Platoon of Autonomous Vehicles", CDC 2023.

[C2]. Vivek Pandey, Guangyi Liu, Arash Amini, and Nader Motee. "Quantification of Distributionally Robust Risk of Cascade of Failures in Platoon of Vehicles.", CDC 2023.

[C1]. Arash Amini, Guangyi Liu, and Nader Motee. "Robust Learning of Recurrent Neural Networks in Presence of Exogenous Noise.", The 60th IEEE conference on Decision and Control (CDC), December 13-15, 2021, Austin, Texas, USA.

Preprints:

[P4]. Guangyi Liu, Arash Amini, and Nader Motee. "Risk of Misclassification in Recurrent Neural Networks."

[P3]. Guangyi Liu, Arash Amini, Martin Takáč, and Nader Motee. "Robustness Analysis of Classification Using Recurrent Neural Networks with Perturbed Sequential Input.".

[P2]. Guangyi Liu, Arash Amini, Martin Takáč, Héctor Muñoz-Avila, and Nader Motee. "Distributed Map Classification using Local Observations" arXiv preprint arXiv:2012.10480 (2020).

[P1]. Mousavi, Hossein K., Guangyi Liu, Weihang Yuan, Martin Takáč, Héctor Muñoz-Avila, and Nader Motee. "A layered architecture for active perception: Image classification using deep reinforcement learning." arXiv preprint arXiv:1909.09705 (2019).

Posters and Extended Abstracts:

Guangyi Liu, Christoforos Somarakis, and Nader Motee. "Origins of Risk of Cascading Collisions in Platoon of Autonomous Cars.", (Extended Abstracts) Workshop on Safe and Reliable Robot Autonomy under Uncertainty, at IEEE ICRA 2022, May 27, 2022, Philadelphia, PA, USA.
Posters of "Origins of Risk of Cascading Collisions in Platoon of Autonomous Cars." and "Risk of Cascading Failures in Multi-agent Rendezvous with Communication Time Delay." at Robot Versatility Workshop, Lehigh University, May 27, 2022, Bethlehem, PA, USA.
Guangyi Liu, Arash Amini, Martin Takáč, and Nader Motee. "Robust Classification with Localized Observations Using Stable Recurrent Neural Networks." Modeling and Optimization: Theory and Applications (MOPTA 2021), August 2021, Bethlehem, USA.

Selected Talks and Seminars:

I gave a talk at NERC 2024 about our paper "Beyond Uncertainty: Risk-Aware Active View Acquisition for Safe Robot Navigation and 3D Scene Understanding with FisherRF'." Sept 28, 2024, Amherst, MA.
I gave a talk on "Towards Safe Learning and Perception-based Networked Control Systems: A Risk-aware Approach" as the GRASP seminar at Upenn. May 24, 2024, Philadelphia, PA.
I gave a talk on "Distributionally Robust Mitigation of Cascading Risk in Cooperative and Competitive Vehicle Platooning" at the xLab, University of Pennsylvania. October 20, 2023, Philadelphia, PA.
I gave a talk on "Risk of Misperception in Decision-Making and Control" on behalf of my advisor Prof. Nader Motee, at the ONR Science of Autonomy Program Review, August 3 2023, Alexandria, VA, USA.

Report abuse