Sandip Das

Student Details:

• Pursuing BS (4 year) in Mathematics and Computing
• Member of the Software Team
• A Sophomore as of Aug 2024

Inverse Reinforcement Learning

Ongoing
Smaller projects done till now for understanding the basics:

• Implemented an Optimal Policy generator by iteratively solving the Bellman Equation for a 2D grid environment, specifically one where each square is either an obstacle or is one with a reward for it. Link

Minimum Curvature trajectory planning

Completed
Reference Paper: Minimum curvature trajectory planning and control for an autonomous race car

• Link to Implementation: AGV-2023-Planning-Teams/minimum_curvature_planner
• Calculated the QP problem that needs to be passed to a solver
• Coded a setup to visualize output trajectories
• Created a Powerpoint presentation to describe the algorithm

Autoware Course

Ongoing
Course Link: Autoware course hosted by Apex.AI

• Documentation Link: GitHub link to my documentation

Long Term Simulation Task

Completed
Task Link: tharun-selvam/sim_task_agv

• This project consists of multiple GitHub repositories edited by myself.
• github.com/real-Sandip-Das/sim_task_agv (Main repository, contains explanations, instructions and links to other packages)
• real-Sandip-Das/aruco_opencv_to_cartographer_landmark (a custom ROS Package for translating Aruco Detection messages to a format useful for Google Cartographer SLAM Nodes)
• I was required to detect Aruco markers from the camera's topic, but initially I couldn't find a package that was able to do that, so I decided to select a specific package, clone its source locally and debug it.
• fictionlab/ros_aruco_opencv (I debugged its noetic branch and made a Bug Fix in this repository: Pull Request for my Bug Fix and then used this in my project)
• Docker: After completing the project, I uploaded the Docker Image for an easier experience of potential evaluators: realsandipdas/sim_task_agv/

Conflict Based Search

Edited and Improved
Link to Paper: Conflict-based search for optimal multi-agent pathfinding

• Task based on implementation of an algorithm described in a paper
• Although this task was assigned during the selection process, I have been polishing my solution (simply, so that I can show this to other people as a project in general)
• GitHub Repository: real-Sandip-Das/Conflict-Based-Search/tree/converting_in_cpp

Introductory ROS Task 1

Completed
Google Doc with Problem statement and relevant details

• In short, contains a publisher and a subscriber node, both written in C++, where things like the published dummy string, topic(for publishing/subscribing to), queue size of subscriber node are customizable using ROS Parameters
• GitHub Repository: real-Sandip-Das/pub_sub_comm
• Documentation: contained in the GitHub repository's README.md file

Introductory ROS Task 2

Completed
Google Doc with Problem statement and relevant details

• This task involves a light amount of mathematics, so I've used LaTeX(MathJax) to document the mathematical details (calculations)
• The entire C++ Source code is documented(along with the mathematical details) using Doxygen (proper instructions are given in the repository's README.md)
• GitHub Repository: real-Sandip-Das/noisy_turtle_playback
• Documentation: contained in the GitHub repository's README.md file
• I know I could just use a library function to generate samples of a Bi-variate Normal Distribution, but I thought it would be fun to involve mathematics with proper documentation in this task.

Introductory ROS Course by ETH, Zurich

Completed
Course Link: Programming for Robotics - ROS

• Documentation: real-Sandip-Das/AGV-ROS-task-doc
• GitHub Repositories: smb_highlevel_controller, smb_highlevel_controller/tree/Exercise2(branch)

Introductory Git Task

Completed
Task Link: Cath3dr4l/AGV-git-task

• Documentation: real-Sandip-Das/AGV-git-task-doc
• GitHub Repository: real-Sandip-Das/AGV-git-task