Mohammed Azharudeen

Developed a simulation environment in MuJoCo, supporting both state- and vision-based inputs for a peg-in-hole assembly task for Learning from Demonstrations(LfD) and Human-in-the-Loop RL. Implemented and tuned various controller modules; Integrated teleoperation modules for spacemouse control, collected demonstration data and explored synthetic data generation. Built a real robot manipulator interface with Flask–ROS server for command execution with actor–learner asynchronous nodes.

Extended codebase of a unified C++ interface for various quadratic programming (QP) to nonlinear programming (NLP) solvers and included coupled tools for MPC, NMPC, SQP, auto differentiation, Hessian regularization methods, polynomial interpolation, collocation methods, etc with appropriate tests. Involved in the Fernbin 2 project, where my tasks included developing simple simulation for testing control systems in Python and ROS2 for ship-vessels/ferry that react dynamically to the surroundings.

Integrated a particle-based contact model to simulate the complex behaviors of granular materials in NVIDIA Isaac Sim, developed an adaptable and robust controller using Reinforcement Learning (PPO) in PyTorch for quadruped (Unitree A1) to follow a unified policy in diverse natural environments and evaluated the performance with sim-to-real additions. Preliminary work presented at 1st German Robotics Conference (GRC2025), and a related paper is being finalized.

Enhanced trajectory planning in autonomous vehicles using Model Predictive Control (MPC).

Studied predominant microscale phenomena subjected to acoustic fields and published results in various reputed journals. Development of a low cost Microfluid Device for Loop-Mediated Isothermal Amplification of DNA.

major in Mechanical Engineering(Specilization: Computational Techniques) and minor in Electronics and Communication Engineering