Developing perception, SLAM, computer vision, deep learning, and sensor calibration solutions for autonomous off-road agricultural machinery operating in perceptually challenging and GPS-denied environments.

FLUENT Robotics Lab:

• Worked on enhancing Human-robot collaborative transportation tasks such as lifting and transporting objects with users, improving human experience with robots in the workspace.

• Built the autonomy software stack for a mobile manipulator, enabling complex tasks around humans, ensuring safety and stability. Contributed to system integration and designed pipelines to characterize hardware and software performance.

• Devoloped a navigation stack for a mobile manipulator in ROS2 with RGB-D feature based SLAM and MPPI (Optimal trajectory control for mobile base), enabling the robot to navigate autonomously in a dynamic indoor setting.

• Integrated multi-modal perception capabilities with computationally limited hardware, leveraging pre-trained models for visual-language manipulation and pose estimation improving user experience in collaborative tasks.

Teaching assistant for EECS 504 Computer vision course at University of Michigan Ann arbor

Responsibilities included:

• Grading student homework assignments and projects.

• Conducting office hours discussing coursework and assignments.

BIRDS (Biologically Inspired Robotics and Dynamical Systems) Lab :

• Deployed a perception software stack using C++ and ROS for a Hexapedal picking robot, using a Intel realsense L515 sensor to detect, track and create a map of targets, while simultaneously performing localization.

• Programmed a computer vision pipeline for target detection and instance segmentation, utilized an early fusion model to integrate with a 3D LIDAR graph SLAM system seamlessly, enabling real-time state estimation and semantic mapping.

Distributed Aerospace Systems and Control Laboratory :

• Presented a practical analysis of an Adaptive Safety-Critical control system that ensures real-time safety and stability on digital platforms with discrete-time updates, comparing it to traditional periodic controllers.

• Designed Self-triggered Exponential-CBF controller in MATLAB that uses Quadratic Program (QP) optimization applied in Zero-Order Hold manner with a notion of Safe period for Higher order systems, resulting in safety and stability in safety-critical applications, overcoming the limitations of periodic controllers.