About Me

Hello, I am Sandeep Thalapanane and I enjoy designing and developing robotic systems. Right now, I am pursuing a Master’s in Robotics at the University of Maryland, College Park, USA. In 2020, I earned my bachelor’s degree in Mechanical Engineering from Anna University, India.

Motivation - “What we know is a drop, what we don’t know is an ocean.”

Where I’ve worked

TRC Robotics

• Robotics Software Engineer Intern (May 2023 to August 2023)

  • Developed a user-friendly Grafana plug-in utilizing roslib JavaScript to facilitate robot control through a web page interface. Created and customized Docker containers to successfully integrate required modifications.
  • Worked on enhancing the Grafana plugin by incorporating features for scheduling, visualization, and robot simulation directly on the web page.

Flextronics

• Senior Analyst (April 2021 to August 2022)

  • Performed multiple daily procurement functions such as handling purchase orders over $6 million worth of electronic parts during covid critical times, efficiently met inventory targets of 1 million USD using BAAN ERP software, and coordinated with vendors and suppliers from all parts of the world to meet the desired targets.
  • Augmented understanding of the procurement process & supply chain management.

Research I’ve worked on

Driving style conditioned for autonomous vehicles

• Gamma lab UMD, Advisor: Dr. Ming C Lin (August 2023 to present)

  • Designed a novel Virtual Reality driving simulator using Unity, SUMO, RoadRunner and will be conducting a study using meta quest VR to collect personalized driving data on different parameters including steering angle, throttle, and timing the brakes, and train a model to learn the driving behavior based on the parameters.
  • Developing a personality classification LSTM model architecture and trajectory prediction algorithm design to predict future trajectory based on the driving style of nearby vehicles in mixed autonomy environments.

Research Assistant

• Multiscale Measurements lab UMD, Advisor: Dr. Hugh Bruck (August 2023 to present)

  • Designed a novel Virtual Reality driving simulator using Unity, SUMO, RoadRunner and will be conducting a study using meta quest VR to collect personalized driving data on different parameters including steering angle, throttle, and timing the brakes, and train a model to learn the driving behavior based on the parameters.
  • Developing a personality classification LSTM model architecture and trajectory prediction algorithm design to predict future trajectory based on the driving style of nearby vehicles in mixed autonomy environments.

Projects I’ve Done

3D Inspection Using NeuS

• November to December 2023
  • Developed and implemented an innovative 3D inspection pipeline using HF-NeuS and DeepCrack for crack segmentation and surface reconstruction.
  • Leveraged HF-NeuS’s ability to capture intricate geometries and handle challenging areas to enable inspection of complex structures.
  • Utilized DeepCrack’s pixel-wise crack segmentation capabilities to accurately identify and analyze cracks in surfaces.
  • Improved the original HF-NeuS model by adding two custom loss functions, resulting in a 25 percent reduction in training time compared to the conventional model.
  • Github repository

Implementation of Robot Motion Planning in Learned Latent Spaces - L2RRT

• April to May 2023
  • Implementing a novel sampling-based motion planning algorithm called Learned Latent Rapidly-Exploring Random Trees (L2RRT) to address the challenges of traditional sampling-based methods in high-dimensional robot motion planning problems.
  • Leveraging a deep neural network to project the problem into a lower-dimensional latent space, where it can be solved more efficiently and then mapping the solution back to the original space to provide a trajectory for the robot to follow.
  • Github repository

Underwater image restoration

• April to May 2023
  • Developed an innovative project aimed at improving visibility in underwater imagery by employing an image formation model to counteract the light scattering caused by water & restore the true colors of the scene.
  • Adapted image restoration techniques from dehazing applications to underwater conditions, using a deep learning model to accurately estimate the depth of objects , thereby improving the efficiency and reliability of the image formation model over conventional methods.
  • Github repository

Virtual car driving using hand gestures

• April to May 2023
  • Inspired by NVIDIA’s virtual driving initiative, developed a remote-controlled automotive system using computer vision & deep learning.
  • Trained a TensorFlow model and translated hand gestures into commands for a robot, simulating vehicle actions like acceleration, braking, reversing, and speed adjustments.
  • Github repository

Implementation of A-star algorithm for path planning of Turtle Bot using ROS2

• April 2023
  • Developed an implementation of A* algorithm in Python for path planning of TurtleBot in a map with obstacles. Utilized ROS2 framework to simulate the robot’s movement in the obstacle space.
  • Integrated the A* algorithm with the ROS2 simulation to generate optimal paths for the TurtleBot to follow and visualize in Gazebo.
  • Github repository

Camera pose estimation by using own Homography and Hough transform functions

• March 2023

  • Developed an algorithm to compute the rotation and translation between a camera and a coordinate frame placed on a sheet of paper, without relying on OpenCV’s built-in Homography and Hough transform.

  • Implemented computer vision techniques such as image thresholding, edge detection, and contour detection to extract the coordinates of the corner points of the sheet of paper.

  • Github repository

Implementation of the Dijkstra Algorithm for path planning of a Point Robot

• March 2023

  • Developed an implementation of the Dijkstra algorithm in Python to find the shortest path between two points in a map with obstacles, using a graph representation of the map and utilizing data structures such as priority queues to efficiently store and process the graph data.
  • Integrated the algorithm with a Pygame animation to visualize the resulting explored nodes and optimal path. Also enabled users to select start and goal points.

  • Demonstrated a deep understanding of fundamental concepts in computer science and mathematics, including graph theory, algorithm analysis, and data structures.

  • Github repository

Implementation of the Breadth First Search algorithm for solving 8 puzzle problem

• March 2023

  • Developed an implementation of the BFS (Breadth First Search) algorithm in Python to find the shortest path between an initial state and a goal state in the 8 puzzle problem, which involves sliding numbered tiles in a 3x3 grid
  • Integrated the algorithm with another python script to visualize the shortest path in the terminal by backtracking from goal state to initial stae.

  • Demonstrated proficiency in algorithms and data structures, as well as familiarity with Python.

  • Github repository

Ball tracking and finding trajectory using OpenCV

• March 2023

  • Developed a computer vision algorithm to track a ball in a video stream, using techniques such as color segmentation.
  • Used RANSAC (Random Sample Consensus) to robustly fit a curve to the ball’s trajectory by implementing least squares, even in the presence of noise, outliers, or occlusions.
  • Github repository

Modeling of a High speed camera mobile manipulator

• October 2022 to December 2022

  • Designed a High speed camera robot arm model with a mobile platform and formulated the Kinematics and Dynamics of the robot links.
  • Built a ROS package and simulated in a custom made Gazebo environment and visualized camera attached at the end effector using RViz.
  • Developed a python program to operate robot arms, camera and reach the desired position by moving the robot using a mobile platform and planned trajectories using MoveIt.
  • Github repository

Implementing the Wall following Algorithm to drive a robot to a goal

• November 2022

  • Formulated the structure of the C++ program using OOP, pointers, and classes.
  • Wrote two types of wall following algorithm using left hand rule and right hand rule to reach a goal in a maze if the goal is adjacent to one of the outer walls of the maze.
  • Programmed to turn around the robot, when it reaches dead ends in the maze and continue to follow the algorithm.
  • Github repository

Designing LQR and LQG controller for a crane system

• December 2022

  • Calculated the equations of motion for the system using the Lagrangian method and the corresponding nonlinear state-space representation.
  • Linearized the system around the equilibrium point and designed an LQR controller for the linearized system and certified the stability using Lyapunov indirect method.
  • Designed an LQG controller for the nonlinear system and obtained the best Luenberger observer for each one of the output vectors.
  • Github repository

Performance and emission analysis on a CI engine using soapnut biodiesel by varying injection timing

• January 2020 to Septemnber 2020

  • Extracted soapnut oil from soapnut seeds and reduced viscosity of soapnut oil by using Transesterification.
  • Tested blended soapnut oil in CI Engine by varying injection timing.
  • Analyzed the results and inferred that the change in performance characteristics was not quite significant, whereas the variation in the emission characteristics was considerable.
  • Concluded that the soap nut biodiesel is more advantageous when compared to the diesel in terms of availability, eco friendly, price and ease of use.

Design and fabrication of a PCM integrated solar still

• August 2019 to December 2019

  • Designed & fabricated a solar desalination system using water as PCM in copper tubes and glass wool as insulation material.
  • The experiment was conducted throughout the day in the sunlight and measured the productivity of solar still on an hourly basis at the night.
  • It was evident from the results that, when water was used as PCM by storing in the copper tubes, the yield increased from 10 percent to 24 percent.

Skills

Technologies	Developer tools	Programming Languages
ROS	Git	C++
Gazebo	Google collab	Python
RViz	VS code	MATLAB
Solidworks	Pycharm
MoveIt
BAAN(ERP)
Microsoft Excel