Projects

I develop algorithms for control and analysis of robots and embodied AI. My work bridges theory and practice, focusing on data efficiency, computational efficiency, and human-robot interaction. Here are some highlights from my research and projects.

Featured Projects

Dynamics-Aware Diffusion Models for Planning and Control

Current Research | 2023 -

Developing diffusion models that understand system dynamics for improved robotic planning and control. This work addresses the challenge of incorporating physical constraints and system behavior into generative models for robotics applications.

Key Contributions:

• Improved planning accuracy for robotic systems, requires less post processing of generated trajectories
• Easily incorporate safety constraints
• Theoretical guarantees for stability and performance

Publications: 📄 Paper 1 , 📄 Paper 2 , 📄 Paper 3 , Code

Intention Inference in Human-Robot Collaborative Environments

Honda Research Institute Internship | 2024

📄 Project webpage

(To appear at IROS ‘25) Developed methods to probe and understand emobodied AI agents intentions for safer human-robot collaboration in automotive environments. This project focuses on real-time intention recognition and predictive modeling.

Key Contributions:

• Improved safety in human-robot collaborative scenarios
• 44% smoother trajectories
• Real-time intention recognition
• Applied to autonomous vehicle decision-making

Data-Driven Control of Nonlinear Systems

*Ongoing Research | 2022 - *

I combine Koopman operator theory and feedback linearization to develop interpretable data-driven controllers.

Key Contributions:

• Interpretable data-driven controller
• Theoretical results on data-requirement for optimal controller

Publications: 📄 Paper 1 , 📄 Paper 2 , 📄 Paper 3

Coverage Control of Ground Robots

Developed distributed algorithms for optimal area coverage using teams of ground robots. Used a optitrack mocap system to track ground robots, and used a consensus protocol for oscillators for controller.

Distributed Intersection Management for Connected and Autonomous Vehicles

Designed a data-guided distributed control algorithm for the safe and efficient intersection management in connected vehicle environments.

Publication: Paper

Online Learning with Heterogenous Sensor Array

Investigated online learning strategies with heterogenous sensors with data arriving at different rates

Publications: 📄 Paper

Additional Projects

Data-Driven Attack Detection in Dynamical Systems

Developed methods to detect and mitigate attacks on dynamical systems using behavioral analysis tools from dynamical systems. Established a tradeoff between system identification and direct attack detection.

Publications: 📄 Paper