Embedded Files
Our Research
Our Research
The backbone of our research program employs applied probability, statistical physics, optimization, linear algebra, and dynamical systems theory to address questions related to three classes of problems:
AI and AV security: We examine the vulnerability of deep neural networks (DNNs) used in AV control and develop defenses against these attacks.
The statistical physics of traffic: The central theme of this work is the development of scaling laws to disentangle transient, short-lived fluctuations from the underlying long-term dynamical processes governing traffic behavior, providing a robust mathematical framework for representing uncertainty in traffic systems.
Neural operators for traffic: We develop novel operators to solve both forward and backward problems in traffic flow.
Below are some examples of our research work. A complete list of our research products can be found here
Below are some examples of our research work. A complete list of our research products can be found here
Adversarial training with general p norms
Adversarial training with general p norms
Traffic Forecasting As A Matrix Completion Problem
Traffic Forecasting As A Matrix Completion Problem
Position-Weighted Backpressure Intersection Control
Position-Weighted Backpressure Intersection Control
Position-Weighted Backpressure
Position-Weighted Backpressure
Standard Backpressure
Standard Backpressure
Stochastic Lagrangian Traffic Dynamics
Stochastic Lagrangian Traffic Dynamics
Sparse Estimation of Travel Times from Streaming Data
Sparse Estimation of Travel Times from Streaming Data
Node Modeling for Urban Networks
Node Modeling for Urban Networks
Probabilistic Fundamental Relations
Probabilistic Fundamental Relations
Incident Localization and Sensor Placement
Incident Localization and Sensor Placement
Report abuse