Rahul Choudhary

Researcher in Robotics and Deep Reinforcement Learning


About Me

Researcher in Robotics and Deep Reinforcement Learning

I focus on the intersection between control in robotics and reinforcement learning, with the aim of developing algorithms and techniques that can endow machines with the ability to autonomously acquire the skills for executing complex tasks.

Name: Rahul Choudhary
Experience: 1.5 Years
Birthday: 1 December 2000
Current Temporary Phone: +1 (514) 430-2608
Permanent Phone: +91 9079102623
Email: rcstark3614@gmail.com


Education & Expericence

My Education

B.Tech in Aerospace Engineering with Micro Specialization in Artificial Intelligence and Applications

Indian Institute Of Technology Kharagpur | 2019 - Present


Class XII

MDS SCHOOL | 2016 - 2018


Class X



My Expericence

Mitacs Research Intern

Mila - Quebec AI Institute | 2022 - Present

Work focuses on Optimisation and Reinforcement learning

Reinforcement Learning Research Intern

Symbiosis Centre for Applied Artificial Intelligence | 2020 - Present

Work focuses on Robotics, Reinforcement learning and Allied areas.

Software Development Intern |

Jan Elaaj Healthcare (P) Ltd. | 2021 - Present

Work focus on using image/signal processing techniques from computer vision to develop health care product.


My Projects

Deep Reinforcement Learning Notebook Series

06 2020 - 04 2021

Implemented basic Deep Reinforcement Learning algorithms like REINFORCE, SARSA, DQN, A2C to advanced algorithms like DQN with Prioritised Experience Replay and target networks, PPO extending Actor- Critic Algorithm, SAC,A3C, Noisy Nets,Rainbow in various easy and medium level difficult gym environments.
Codes were written from scratch in tensorflow framework in python.Reproduced approximately 70-80 of the level of accuracy provided in the research papers that introduced the algorithms.

Robot Manipulation Task using IRL and SAC (Learning from Human Demonstration

10 2020 - 11 2020

Implemented Time-Contrastive Networks and Soft-Actor Critic to form a pipeline of inverse reinforcement learning to perform robotic grasping.
The robot is able to learn to pick up objects similar in shape but different in size from only one video demonstration of human picking up the block.

Robotic Car Path Planning

06 2021 - 06 2021

Utilized the Twiddle algorithm to optimise parameters for a PD Controller and combined with a Smoother to make a robotic car reach its goal on a smooth path without colliding.
Implemented a particle filter for localisation of the car and A* Algorithm to find optimal path.