Anushtup Nandy

My current focus: deploying QP-based MPC with ESDF collision checking that hits 0.02cm positioning accuracy at millisecond solve times. on real hardware, with sub-microsecond IPC latency between the C++ control loop and Python planning layer. Building and refining the robot’s tension and contact dynamics model- characterizing how forces propagate through a hyper-redundant structure to tighten the gap between predicted and actual behavior on hardware.

Project 1: Developing a comprehensive full-scale dynamic model for the 7-DOF Yomi Robot, critical for optimizing its performance, predicting behavior, and refining control algorithms: - Researching different methodologies for accurately modeling the complex coupled dynamics inherent in high DoF robotic systems. - Applying the Recursive Newton-Euler Algorithm to solve for the robot's dynamics, complemented by System Identification techniques to precisely extract embedded inertial parameters. - Exploring Neural Operator architectures (Laplace Neural Operator and Complex Neural Operator) as an innovative approach to learn intricate robot dynamics. Focused on achieving an interpretable and physics-constrained understanding, opening avenues for data-driven modeling with physical insights. - Integrated Pinocchio based System ID, robot model, Dynamics and Torque control into Software stack and demonstrated gravity compensation capabilities. Therefore, providing insight into reducing the cost of 1 YOMI robot unit by 11.6% through removal of F/T Sensor. Project 2: Designed custom fixture - Designed fixture, in SolidWorks, for holding 4 PCBs of tracker arm to enable smoother flashing - To increase ease of usage, added drafted columns with shelfs and cover plate for PCBs - Soldered pogo-pins to enable faster flashing without having to pick up boards.

TAing for course of Mechanics: - checking assignments - setting questions - taking office hours to answer doubts of students

TAing for the course of Artificial Intelligence: - Answer students' questions and clarify doubts related to AI concepts, assignments, and projects -Lead recitation sessions to supplement lectures, providing additional examples, explanations, and practice problems - Evaluate assignments, quizzes, and exams, providing constructive feedback to students - Hold regular office hours to meet with students, address questions, and discuss course material - Foster a supportive learning environment, encouraging students to explore AI concepts and ask questions

- Evaluating homework, projects and exams of over 90 students - Answering doubts during office hours

Working on multiple projects spanning experimental fluid mechanics, image processing, mechanism design and experimental bio-inspired robotics at HM Lab. PROJECTS: - 3D reconstruction of a Flapping Mechanism : frame by frame reconstruction of flapping structure under cross flow from a variable flow wind tunnel. Used Python (Open3D, OpenCV) and 2 intel realsense D435i depth cameras. - Bio-Inspired Undulatory Fin-based Underwater Propulsion System (*on-going*) : developed a modular undulating mechanism capable of propulsion under-water using servo-motors, arduino and silicone flap. Studied propulsive force produced using single-axis load cells of 1kgf max value and Ni-DAQ modules. Studied wave formed using PTC high speed camera and matlab. - Flapping and pitching wing mechanism for Bio-inspired UAV : Constructed a flapping and pitching mechanism using 2 servomotors and studyed its behaviour under cross flow from a wind-tunnel.

Worked on 2 projects spanning motion planning, mechanism design, image processing and deep reinforcement learning in Robotics Labs at Bits Hyderabad. PROJECTS: - Path planning using Deep RL : Path planning of mobile robot in office environment using DQN (value based deep reinforcement learning algorithm) under time constraints of 600 and800 seconds. - Single Legged Hopper Robot : PID control of single legged hopper robot. Simulation using PyBullet.

Developed algorithms for multi-agent path finding (MAPF)problem called "path finding with refuelling" (PF-RF).

- Innovated a high-efficiency injection holder and nozzle mechanism, reducing the size by 20%. - Developed FSR based signalling system to help protect syringe from breaking under pressure. - Led a team to source local materials, exemplifying leadership and collaboration skills.