projects

work

RL with the Franka Robotic Arm

I used Reinforcement Learning to train a SAC-based RL agent to control the Franka Robotic arm visually.

Hybrid Local Planner

I implemented a ROS Local Planner to navigate a mobile robot in this project. I used Dijkstra's algorithm and the Hybrid A star algorithm for the planner. I used TurtleBot3 to test the algorithm.

fun

TGR

TGR is a Raspberry Pi (4B) based 4wd robot created for real-world ML/RL-based projects. The robot can perform the line-following task using the camera image, and the robot can be controlled using an Android App. I plan to implement projects involving ML and RL using the robot.

Meteors and Deep Q-Learning

In this project, I created an RL agent that is able to play the game, Meteors. I used the python Arcade library to create the game from scratch. I used Deep Q-Learning to train the RL agent in this project. I experimented with various improvements of DQN and tested how those improvements worked out in playing Meteors.

Project Tennis

In this project, I experimented with the Unity ML-Agents Tennis environment. I used the MADDPG algorithm in my project. MADDPG is a variant of the DDPG algorithm for multi-agent environments.