The Journey So Far

Exploring control with MuJoCo and Brax by training PPO in multi-environment setups. Currently studying quadruped dynamics, focusing on stabilizing the robot in-place.

  • training PPO using Brax.
  • experimenting with controlling and balancing a quadruped model(unitree go2).

Went deep into set computation and reachability using Julia. Discovered mathematical structures like sparse polynomial zonotopes and zonotopes, reading papers and implementing core functions.
some favs - SPZs,Ellipsotopes

  • studied and experimented with reachability analysis and set representation.
  • implemented order reduction and set enclosures based on research papers. SPZs
  • mathematical rigor meets control theory.
  • contributions merged in LazySets.jl— GitHub

This project was my first exposure to actually working with LLMs. Implemented Retrieval-Augmented Generation (RAG) with Qwen2.5-Instruct to chat with BeagleBoard documentation

  • built a RAG tailored to BeagleBoard documentation.
  • BeagleMind

Worked on a 3-DOF robotic arm project with ROS2 and Gazebo. Interfaced with microcontrollers via micro-ROS.

  • forward and inverse kinematics implemented.
  • simulation with ROS2 Humble and Gazebo.
  • communication using micro-ROS.

This was the start of my journey in controls

  • Designed and implemented an LQR controller in ESP-IDF
  • Stepper motor control with FreeRTOS on ESP32 and MPU6050 feedback.
  • Simulated and tuned system parameters in MATLAB.

Read the blog post | View Project Page

Participated in the e-Yantra Robotics Competition (eYRC), working on autonomous drone navigation and vision tasks.

  • Implemented drone control and A* path planning in ROS2.
  • Used OpenCV for Aruco Marker-based object detection.
  • Reached top 30 out of 600+ teams; advanced to Stage 2.
  • Fixed and revived an old 3D printer — Read the blog
  • Taught Control (LQR, MPC), OpenCV, and Mechanical Design to juniors.
  • Conducted workshops and study sessions on Control and Robotics.

Awe for how mathematical models can describe and influence physical systems.
Projects in mind — from theory to hardware.

  • Hoping to bridge Julia-based computation with real-world implementation.
  • MPPI.
  • MORE RL MORE MODELS.