Below are videos highlighting a few threads of our research. Additional videos can be found on my YouTube channel. A complete list of publications can be found on my Google Scholar profile.
Semantically Safe Control
For robots to safely interact with people and the real world, they need the capability to not only perceive but also understand their surroundings in a semantically meaningful way (i.e., understanding implications or pertinent properties associated with the objects in the scene). In this work, we aim to close the perception-action loop and develop a perceptive safety filter framework that allows robots to incorporate a semantic understanding of their operating environment and comply with “common sense” safety constraints beyond collision avoidance. RA-L 2025, ICRA 2024
Perceptive Safety Filter
When robots navigate in unstructured environments, they must perceive and construct a map of their surroundings to plan their path and avoid collisions. In this work, we introduce a perceptive safety filter that combines Control Barrier Functions (CBFs) with Visual-Inertial Simultaneous Localization and Mapping (VI-SLAM) and dense 3D occupancy mapping to ensure safe navigation in complex and unstructured settings. Our system relies entirely on onboard IMU measurements, stereo infrared images, and depth images. It autonomously corrects teleoperated inputs when they are deemed unsafe, enhancing safety during navigation. ICRA 2024
Multiagent Coordination
Distributed Model Predictive Control (DMPC) offers an efficient and scalable approach for coordinating a swarm of robots; however, solving the DMPC problem is challenging due to (non-convex) quadratic constraints. Our work proposes an alternating minimization approach that enables a swarm of drones to safely navigate through cluttered and unstructured environments. ICRA 2023, ICRA 2024
Drone Swarm Choreography
Drone performances are making their mark in the entertainment industry. Designing smooth and safe choreographies for drone swarms often requires expert domain knowledge. How do we facilitate this process? In this work, we introduce a language-based choreographer that integrates the reasoning capabilities of large language models (LLMs) with a Distributed Model Predictive Control (DMPC) framework to facilitate the design of deployable drone swarm choreographies.
Safety Filter Fundamentals
Control barrier certification is a common framework for designing safety filters. These safety filters are intuitively used to protect robot systems by minimally adjusting potentially harmful inputs before they are applied. In this work, we investigate fundamental problems related to control barrier certification. ACC 2024, L-CSS 2023
Doctoral Seminar: Neural Networks as Add-on Modules for Improving Robot Performance
This is my Doctoral Seminar talk summarizing 5 years of Ph.D. research in 20 minutes.
Bridging the Model-Reality Gap with Lipschitz Network Adaptation
Model-based approaches have proven to be very efficient and effective when the dynamics is well-characterized. In this work, we explore how learning can be used to bridge the model-reality gap for uncertain systems to empower model-based design techniques. RA-L 2021
Characterizing System Similarity for Experience Transfer
Various knowledge transfer approaches have been proposed to leverage experience from a source task or robot—whether real or virtual—to accelerate learning in a new task or robot. However, inappropriate knowledge transfer can lead to negative transfer or unsafe behavior. In this work, we investigate system similarity and introduce a data-efficient algorithm for estimating the similarity between pairs of robot systems, enabling safer and more effective knowledge transfer. ICRA 2020, L-CSS 2020
Knowledge Transfer with Online Learning
A team of robots can often achieve more than single robots; however, training each individual robot in a team can sometimes be non-economical. If we are given a model trained on one robot, can we leverage the learned experience to enhance another robot? How do we characterize the similarity between robots, and what is the implication of having a higher similarity in the knowledge transfer problem? ECC 2019
Neural Networks for Quadrotor Impromptu Tracking
While it may be easy to design a controller to stabilize a system, tuning the controller parameters to achieve good tracking performance can be hard. Can we design a DNN add-on module to enhance the tracking performance of black-box control systems on arbitrary trajectories? CDC 2017, IJRR 2020