I completed my MS in Robotics from Worcester Polytechnic Institute (WPI) in May 2023.

I received my undergraduate degree from Birla Institute of Technology & Science, Pilani (India), where I worked with Prof. B. K. Rout at The Centre for Robotics & Intelligent Systems(CRIS).

My research interest lies within the areas of Deep Learning & Computer Vision.

Sections

• Skills & Interests
• Work Experience
• Research Assistant
• Relevant Projects
• Innovation Disclosures & Patents
• Conferences & Proceedings
• Activities & Achievements
• Teaching Experiences

Skills & Interests

Perception, Computer Vision, Sensor Fusion, Deep Learning, SLAM, NeRF, Depth and Pose Estimation

Work Experience

Optimal scan positioning for an automated scanning process with an autonomous robot platform

Explored state-of-the-art techniques in computer vision & deep learning for navigation of aerial robot Athera

Estimated individual states for human model understanding

Worked on the autonomous warehouse robots

Worked closely with autonomous sprayers & tractors for American and European markets

Research Assistant

Working on an automated framework to determine the optimal scan planning for a stop-and-go mapping procedure with a quadruped walking robot based on BIM

Worked in collaboration with Eversource Energy on autonomous robot that can patrol cables for deterring birds from congregating near utility assets

Deployed custom trained YOLO model on Jetson Nano powered robot; integrated camera & Time of Flight sensors using ROS Noetic

Defined a 2D and 3D planner for In-Hand Manipulation Platform on External Surfaces using A* algorithm with Geodesic and Euclidean distances as heuristic

Developed the concept of an Autonomous Bike, with the aim to reduce accidents & achieve better control on uneven terrain

Built simulation model & small scaled prototype withstanding upto +/- 13 degree disturbance using Gyroscope & PID controller

Backed by a renowned Indian electric vehicle manufacturer to develop full-fledged self-balancing electric bike's model

Focused on combining vision-based navigation approaches with the notion of modeling of obstacles and classifying them as per their type, pose and spatial measurements

Evaluated performance of proposed algorithm using criteria like the total number of nodes explored and total path cost for all the cycles

Relevant Projects

Reconstructed a 3D scene from a set of images with different viewpoints using NeRF

Implemented research papers related to GANs: DCGAN, Pix2Pix, Conditional GANs & CycleGAN

Developed a code that reads a depth image and utilizes camera parameters to deproject it into a point cloud, filters out the table from the point cloud, and projects the remaining objects onto a top-down depth map

Extracted features from images using vehicle’s camera setup to find matches, implemented match filtering by thresholding distance

Estimated the camera motion between subsequent photographs using PnP & Essential Matrix Decomposition to build trajectory

Reconstructed 3D image using 10 2D-image sequences of Dinosaur dataset taken from multiple angles with same camera calibration

Developed unified encoder-decoder architecure using PyTorch to perform depth & surface estimation with semantic segmentation

Performed experiments using VGG16 & ResNet versions as encoders with VGG16 offering good trade-off between peformance & speed

Executed Zhang Camera Calibration, utilized eigen decomposition to solve a homogenous system of linear equations and maximum likelihood estimation (MLE) for optimizing calibration parameters

The mean re-projection error post optimization was close to 0.5 pixels

Performed HOG feature extraction on labeled training image set, trained Linear SVM classifier, implemented sliding-window technique, created heatmap to follow detected vehicles and estimated bounding box on detetcted vehicles

Compared results with Yolo

Used max IoU & least false positives as comapring parameters

Stitched images by projecting images onto a cylindrical surface, estimated homography between image pairs using feature correspondences

Used RANSAC for removing the outliers among the feature matches

Created end to end pipeline, used Opencv library for face & eyes detection and wavelet transforms for feature engineering

Built model using SVM, logistic regression & random forest, fine-tuned model using gridsearchcv

F Score: SVM: 84.12%, Random Forest: 70.6%, Logistic regression: 80.95%

Innovation Disclosures & Patents

Conferences & Proceedings

Presented a novel approach to optimize scan planning in 3D indoor spaces using a Building Information Modeling (BIM)-based autonomous scan planning framework with a quadruped walking robot

Proposed a lab prototype of a PID control tuned, self-balancing bike that can be further extended to autonomous bike with path and obstacle detection to travel using global positioning system

Attempted to determine an optimum path, given the start and end points of the end-effector of a virtual delta robot, and to minimize the energy consumed by the motors to drive the platform. In addition, it synthesizes a configuration that consumes the least energy while traversing a given path

Presented a design of an innovative do-it-yourself (DIY) technological system to recycle used absorbent hygiene products and obtain valuable secondary raw materials in a completely eco-friendly manner

Presented a novel concept and working mechanism of spanwise adaptive wings for an unmanned aircraft with retractable configuration for maneuvering in narrow spaces, and extended configuration for enhanced gliding and increased maximum speed as needed

Sole undergraduate team selected to present a paper at the conference, distinguished among seasoned scientists from ISRO

Activities & Achievements

Integrated the robust manipulator with a soft robotic gripper for precise and gentle handling of delicate crops and to deliver exceptional results in autonomous fruit, vegetable, and flower harvesting

Combined computer vision, dynamics, controls, motion planning, and soft robotics to create durable robots capable of withstanding the challenges of outdoor agricultural environments

Won accolades in the AMD Robotics Innovation 2023 Challenge

For details of design, please contact via Email

Developed a Group 3 size unmanned Vertical Take-Off and Landing(VTOL) Aircraft that achieves high-speed forward flight (relative to current VTOL aircraft) and efficient hover through the use of novel reconfigurable propulsive and lifting devices for mega-city type environments

Participated in the American Helicopter Society 35th International Student Design Competition (2018) and was the only Indian team to win this competition since 2011 in the undergraduate category, that too in its first attempt

Teaching Experiences

• ECE 2311 - CONTINUOUS TIME SIGNAL & SYSTEM ANALYSIS (Oct-Dec2021)
• ECE 2312 - DISCRETE TIME SIGNAL & SYSTEM ANALYSIS (Jan-Mar2022)
• ECE 2010 - INTRO TO ELECTRICAL & COMPUTER ENG (Mar-May2022)
• MATH 1022 - CALCULUS II (Aug-Oct2022)
• MME 532 - DIFFERENTIAL EQUATIONS (Jan-Mar2023)
• MME 527 - LINEAR MODELS II (Jan-Mar2023)