Drone Swarm Challenge

Inter-IIT Tech Meet 2023, IIT Kanpur

We transformed an existing ROS-based architecture into a platform-independent and user-friendly Python API to control the drones and perform swarming. Having designed a post-flight analytics dashboard to assess and tune the algorithm using React.js and Plotly, we achieved stable hovering and vision-guided rectangular motion of drones by implementing a multi-axis PID controller.

RL Games Hackathon

National Champions, Shaastra 2022

We created bots using reinforcement learning to compete with other bots in a virtual two-player 2D game setting. With a novel feature engineering technique inspired by the decision tree algorithm, we implemented two Deep RL algorithms namely DQN and Policy Gradient based. In the video, our agent is A (the blue snake) vs and opponent in the tournament.

Mission Planner for Autonomous Robots

Introduction to Robotics, KTH

In this project, I implemented a mission planner on the TIAGo Robot using behaviour trees that achieves autonomous navigation and manipulation based on high level instructions like "Go to Table A, grab the cube and place it on Table B".

For the low level tasks, I implemented an Inverse Kinematics solution for the robotic arm using its Denavit-Hartenberg parameterization, Particle Swarm Optimisation for pose estimation and RRT* for navigation.

Flipkart Grid Robotics Challenge

We implemented a central monitoring/navigation system aimed to control robots in a warehouse. Using OpenCV to understand the arena and estimate the position and orientation of the robot, we developed a central navigation system to instruct the robot on go to the desired location and drop the packages.

Wells Fargo Quantitative AI Hackathon

3rd Place, Shaastra 2022

Secured 3rd place out of 500+ teams registered across the country in the national-level quantitative AI hackathon in which we forecasted the implied volatility surface of options over 60 trading days using 2.5 years of past volatility surface data.