Current research projects

Robust Deep RL for Time-Varying Control (Particle Accelerator Tuning)

This project studies hybrid learning-and-control methods that combine deterministic actor-critic reinforcement learning with bounded extremum seeking for robustness under drift and nonstationarity. I am also exploring safety-aware supervision that detects out-of-distribution operation and determines when to stop trusting the learned policy.

Read the ACC 2026 paper

OOD-Aware Hybrid ES–DRL Control

This project develops Mahalanobis-guided latent out-of-distribution detection for test-time controller switching in nonlinear time-varying systems. A variational autoencoder learns an in-distribution latent representation, and the resulting OOD score determines when to use fast RL control or switch to robust bounded extremum seeking.

Read the ICML RLxF Workshop 2026 paper

Vision Representation Learning for RL

This work builds autoencoder and latent-dynamics pipelines that compress high-dimensional visual observations into compact states for downstream prediction, reinforcement learning, and control. The goal is to improve scalability while preserving task-relevant information.

Space Robotics: Solar Panel Deployment Failure Detection and Repair

I work on vision-based failure detection for satellite solar-panel deployment and on learning-based corrective-control strategies for recovery from jammed or stuck mechanisms.

Prior engineering and research work

Flight-Control Simulation and EFCS Development

At Airbus, I worked on electronic flight-control-system packages for training simulators, malfunction logic, validation workflows, and technical support for control and stability behavior.

UAV Modeling, Estimation, and Control

At Tata Advanced Systems, I worked on nonlinear six-degree-of-freedom aircraft modeling, autopilot development, Total Energy Control System design, extended Kalman filtering, system identification, and flight-performance analysis.

Swarm Robotics and Autonomous Systems

My earlier engineering work included swarm algorithms for quadcopters in disaster-relief scenarios, obstacle avoidance, and sensing pipelines using LiDAR and onboard systems.

Structural Health Monitoring

My undergraduate research studied guided-wave-based online damage detection in aerospace structures using piezoelectric actuators and sensors, finite-element modeling, and experimental validation.