Sanjay Jayalakshmi Prabakar

Working under the guidance of Prof. Milton Aguirre at SEEDS Lab, Purdue .

Build a 3-DOF static balancing manipulator that compensates for gravitational torque.

Designed and built a compliant gripper and an algorithm for estimating the force on the end effector using image feedback from compliance in grasping applications.

Designed and assembled a rigid and compliant prosthetic arm with a custom trigger and Dynamixel motor for control.

purdueseedlab

Worked under the supervision of Professor Nirav Patel.

Developed a mathematical method to apply a software-based Remote Center of Motion (RCM) constraint for Robotic-Assisted Minimally Invasive Surgery (RAMIS) in task space for 5+ DOF manipulators.

For Letter Of Recommendation please mail [Professor Nirav Patel]

Worked under the supervision of Professor Harish P. M at SysIDEA Robotics lab, IITGN on FPGA-based current control for 2-DOF planar manipulator setup, utilizing Iterative Learning Control and PD controller equations to analyze human motor control properties, such as savings, retentions, generalization, interference, and spontaneous reactions.

For Letter of Recommendation , [LOR Link] .

This challenge was orginazied by Freelancer on behalf of NASA - National Aeronautics and Space Administration, in collaboration with Matt Hansen and Ibility LLC, and with support from #ROS2 and Open Robotics community.

I won Honorable mention award and cash price of 400$. [Link to Certificate].

The Package offers a comprehensive simulation of NASA's upcoming CADRE mission, which is an initiative scheduled for Spring 2026. The package includes a highly realistic lunar environment designed in Blender, configured using real-world NASA data. The CAD model of the CADRE rover was developed from scratch in SolidWorks and integrated into ROS 2 Humble. It also supports advanced features like point cloud-based lunar mapping, teleoperation of multiple rovers, and compatibility across various development platforms like WSL and VMware Linux. Additionally, the entire package is open-sourced, including CAD files and Blender assets, allowing for broad accessibility and future development.

The Make-a-thon is an 8 weeks long online challenge bringing together engineers, technologists, lawyers, designers and business professionals to create applications that use AI to improve legal services and access to justice. Participants may use any AI models and can build products on top of the following AI models among others.

It was 8-week roller-coaster journey of over 300 participants prototyping an AI product to improve the justice system during the OpenNyAI Make-a-thon and we Won 3rd place and received a grant of $1800/₹150,000. [grant Link].It is Python-based GUI, which acts as a first-level filter for the officer/institution deciding the application for appointment of an Arbitrator. "

For my Final year project in Btech, I worked on a project titled “Prediction of Health Problems and Recommendation System using Machine Learning and IoT”. Main contribution and the idea was towards creating, GUI based framework where Disease will be predicted based on users symptoms. Due to biased predicted results, we came up with the idea to use 7 ML classification algorithms.

Took part in FLL 2015 and won best team work award in regionals.

My research work at IIT Madras. Accepted to Advances in Robotics 2025

Developed an open-source software stack for integrating KUKA LBR iiwa7/UR5 or other 6-DOF robotic manipulators into a teleoperated surgical robotics system. The system ensures real-time enforcement of the Remote Center of Motion (RCM) constraint while offering intuitive and seamless teleoperation functionality. Will be adding the Link soon for package as soon as paper is published.

Engineered a 2-DOF U-Arm static balancing manipulator that compensates for gravitational torque using couter weights.

Re-Engineered with motors to measure the current with and without weights as shown in the video.

This project uses deep reinforcement learning to teach an agent how to balance a pole on a cart using the Cross-Entropy Method (CEM).

The implementation is built from scratch using PyTorch and the MuJoCo physics engine. It is inspired by the method presented in the book Deep Reinforcement Learning Hands-On by Maxim Lapan.

Elite episodes are selected based on performance, and the policy network is updated using actions from only those successful episodes. This approach enables fast and intuitive learning, making it ideal for simple control tasks like CartPole.

Future work includes implementing more advanced algorithms such as PPO and DDPG for more complex robot environments.

Programming two industrial cobot manipulators to complete a pick-and-place task. Programming autonomous mobile robot (AMR) to handle material transport efficiently. Integrate the cobots and AMR in NodeRED to create a fully automated workflow.

Objective: Developed a ROS-based simulation for NASA's planned CADRE trio mission to map the Moon, scheduled for Spring 2025.

Lunar Landscape Design: Created a lunar environment using Blender, based on publicly available lunar images from NASA’s gallery (https://science.nasa.gov/gallery/moon-images/). Configured world parameters such as gravity, atmosphere, and pressure using data from NASA’s planetary factsheet for gazebo (https://nssdc.gsfc.nasa.gov/planetary/factsheet/moonfact.html).

CAD Model Development: Constructed the CAD model of the CADRE rover from scratch using images and information from NASA JPL’s website (https://www.jpl.nasa.gov/missions/cadre). Implemented the model in SolidWorks and converted it to URDF format for use with ROS2 Humble.

Simulation and Integration: Integrated rover controls, IMU, LRF, and RGBD sensors for lunar mapping within the simulation. Developed Python code to utilize point cloud data for mapping in relation to the Lander.

Teleoperation and Control: Created a Python GUI for teleoperation control of individual rovers, including camera visualization. Set up ROS2 launch files to facilitate spawning multiple rovers as needed.

Cross-Platform Compatibility: Tested the simulation on Windows Subsystem for Linux (WSL) and VMware Linux, ensuring compatibility for developers who do not wish to dual boot or have access to a Linux environment.

Open-sourced : Open-sourced SolidWorks CAD files, Blender files.

A open-source Python ROS2 package to use sampling-based planning algorithms such as Astar RRT, RRT* to control any mobile Robots. will be adding the Link soon for code.

we built a robot from scratch, applying various concepts from both the lab and theory, such as finite state machines, sensors, actuators, filters, and more. Specifically, I worked on the Flywheel launcher, while other team members worked on other subsystems.

Designed and built a compliant gripper and an algorithm for estimating the force on the end effector using image feedback from compliance in grasping applications.