Vineet Kumar Singh

EDUCATION

University of Maryland Master of Engineering in Robotics College Park, 2022 - May 2024

• Cumulative GPA: 4.0/4.0

• Winner of the Final Competition in Path Planning for Autonomous Robots Course. Team of 2 members.

• Worked on 3D vision, Robot perception, planning, and Software Control design for autonomous manufacturing plants.

National Institute of Technology B. Tech in Electrical and Electronics Engineering Jamshedpur, 2016 – 2020

• Cumulative GPA: 8.35/10.0

• Technical Head of Team Phoenix – Aeromodelling Team of NIT Jamshedpur. Competed in SAE Aero design West in 2017, 2018, 2019 and SAE India in 2018, 2019.

• Major Project: Space Vector PWM Controller for 3-phase grid-connected PV System. SKILLS

• Programming Languages: ROS2, bash, Python, C++, PyTorch, PyTorch3D

• Design & Simulation Software: SolidWorks, MATLAB Simulink, Gazebo

• Other Skills: Linux, Data structures, algorithms, robot modeling, perception, planning, 3D vision, control systems.

WORK EXPERIENCE

University of Maryland Teaching Assistant College Park, Aug 2023 – Present

• Spring 2024: Teaching Assistant for “Building a Manufacturing Robotic Software System” course for Graduate Students

(ENPM663).

• Fall 2023: Teaching Assistant for “Introductory Robot Programming” course for Graduate Students (ENPM809Y). ALSTOM Transport Buyer Oct 2020 – Jul 2022

• Successfully developed suppliers of brake system test benches in India (for Pneumatic Brake Valves, AGTU, Brake Caliper & Actuator, and TBU) and did supplier selection, contracts, and price negotiation with 3% YoY productivity. Secured cost avoidance of 1.6 million Euros through this project.

• Localized parking brake release mechanism with Indian suppliers and standardized this part for all upcoming projects. Achieved savings of 150K Euros per project for a total of 3 projects.

• Led the technical non-performance claims settlement with suppliers and published weekly sourcing team reports.

PROJECTS

3D Vision Projects Sept 2023 – Dec 2023

• A comparative study of Neural Implicit Representation based SLAM techniques: Did a comparative study of five prominent neural SLAM methods—iMap, NICE-SLAM, Co-SLAM, Nerf SLAM, and vMAP— which were evaluated for computational efficiency and suitability for resource-constrained robots.

• Single View Image to 3D: Designed a Neural network with an encoder and a decoder to understand the image, and then predict the 3D shape as a voxel, mesh or point cloud.

• Volume Renderer: Implemented a NeRF model from scratch including the random point and ray sampling and the loss function. Implemented an implicit volume as a Multi-Layer Perceptron (MLP) which mapped the 3D position to volume density and color.

• Point Net Implementation: Implemented the Point Net architecture for the classification and segmentation of point clouds. Analyzed the robustness of the model by introducing rotations and changing the number of sampled points.

Advancing Esports Analytics Sept 2023 – Dec 2023

• Project done in a group of 4 students and as part of course work. We designed a deep learning approach to Valorant Match prediction using CNN and LSTM.

• We used CNNs to analyze key in-game visual elements while employing LSTMs to understand the temporal patterns to predict the winner of a game round with accuracy higher than previously possible.

3D reconstruction: Structure from motion Apr 2023 – May 2023

• Designed the algorithm achieving 3D point cloud registration for the object under study.

• Multiple sets of stereo images were used for creating point clouds which were merged to form a single 3D object.

• Built a hardware setup with a turntable and a stereo camera setup using single raspy camera which took multiple stereo images of the object rotating on the turntable. These stereo image sets were used to create 3D point clouds followed by 3D registration to complete the 3D reconstruction.

Camera Calibration, Image transformation, and projections Mar 2023 – Apr 2023

• Implemented the camera calibration using a given set of calibration images. Wrote the code from scratch using fundamental concept of RQ decomposition and projection matrix.

• Worked on homography matrix from scratch and validated the calculated results by stitching multiple images together using SIFT feature matching and calculated homography.

• Implemented Camera pose estimation by reading extracting objects from a given video using Hough transform. Calculate the real time rotation and translation of the camera for every frame of the video. ARIAC 2023 Challenge Feb 2023 – Apr 2023

• Designed a complete software system for a manufacturing robot in a dynamic environment having 2 different robots.

• The Software control system is capable of handling uncertainties/agility challenges which includes handling faulty parts, picking up parts from a moving conveyor belt, and handling insufficient orders.

• The entire code is written in C++ and uses ROS2 and MoveIt for controlling robots and RVIZ for visualization.

Implemented Path Planning algorithms for Non-Holonomic Constraints Mar 2023 – Apr 2023

• Designed a graph generation algorithm with an action set for a differential drive robot with non-holonomic constraint representing real world motion scenario.

• Implemented graph search algorithms of A*, Dijkstra and weighted A* to find the optimal path in generated graph.

• Simulated the result using ROS and Gazebo to ensure the algorithm considers real work scenarios of robot and object clearances.

Trajectory Tracking and Plane fitting for 3D point cloud Feb 2023

• Extracted the trajectory of an object in a video and curve fitted the trajectory using Least Squares method without using any predefined libraries. Implemented using Python, OpenCV, and NumPy.

• Implemented Total Least Squares and RANSAC from scratch for plane fitting for 3D point cloud dataset. KUKA LBR iiwa CAD and Kinematics modeling Nov 2022 – Dec 2022

• Modeled the KUKA LBR iiwa R820 robot from scratch in SolidWorks based on the KUKA specification sheet.

• Installed controllers on the model by modifying the URDF, and ensured all joint constraints are within safety limits.

• Made ROS package for forward and inverse kinematics solver (from scratch) for the model and simulated the pick and place objective in Gazebo.

LQR and LQG controller for double pendulum on gantry crane Nov 2022 – Dec 2022

• Developed dynamic model of system using Euler Lagrange Equations, linearized the system, and analyzed the controllability and observability.

• Implemented LQR and LQG controller and simulated the results in MATLAB for the linearized and the original non-linear system.

Mobile robot control using ROS2 Nov 2022 – Dec 2022

• Designed algorithm for a turtlebot3 to go to a given location and read Aruco marker using a mounted camera, and then go to the final location based on Aruco marker Id. Used TF frames, navigation stack, and dynamic frame broadcaster.

• Packages written in ROS2 Galactic and C++. Simulated the result in Gazebo and camera output using RVIZ.

Space Vector PWM Controller for 3-phase grid-connected PV System Oct 2019 – Apr 2020

• Designed a novel current based controller for integrating the primary grid with a local grid of photovoltaic system. The system focused on reducing the switching transients while shifting the power source from primary to secondary and vice- versa.

• The system uses SVPWM technique for the control and achieved Total Harmonic Distortion (THD) results better than existing methods.

• The complete simulation was done in MATLAB and grid model implemented in Simulink. The results were worked on HIL testbench to check the real-world results that validated the theoretical result.

SAE Aero Design West 2019 Van Nuys, LA, Oct’ 18 – Apr’ 19

• Represented NIT Jamshedpur at the competition and won 1st position in design report and 9th position overall.

• Led the calculations and design team for detail designing in CATIA and SOLIDWORKS and analysis in ANSYS and XFLR5. The model had a final wingspan of approx. 114 inches.

• Fabricated the model (Empty weight-14lbs) and achieved the desired lift values in test flight (Payload lifted-17.50lbs).

• Designed the complete electrical circuit with 1200W BLDC motor, coupled with 112A DC (max.) ESC and a total of 5 servos of various ratings for control surfaces.

OJASS’18, Annual technical fest of NIT Jamshedpur Jan 2018 - Apr 2018

• Won 1st position in Hurdles Hunter. Designed a pick and place robotic arm with DOF 6 mounted on bot. Robot completed the full track of pick and place objectives while travelling though inclined plane and moving track.

Contact this candidate