Machine Learning | Human Machine Interfacing

• Professional gaming is being grown exponentially in recent years and the growth of this industry boomed in COVID-19 era. Gamers buy expansive hardware for competitive gaming just to keep communication delay from hardware to the operating system as low as possible.
• This research is focused on the development of low latency Human-Machine Interface (HMI) using Machine Learning (ML) and Electromyography (EMG) to capture muscle activity from the hand and and predict the mouse clicks. The whole setup consists of three electrodes on the back of your palm, along with ADC connected to the OS. After one training session, the click reaction time improved by an average of 45 ms up to 150 ms.

Reinforcement Learning | Deep Learning

• Worked on a novel algorithm in multiagent Re-enforcement learning (RL) domain, which involves learning tasks from trajectories and experts. This was done at Univeristy of Waterloo with Prof. Mark Crowley and their research team.
• Developed Deep Learning (DL) models for Deep Q-learning from Demonstrations (DQfD), Continuous control with deep reinforcement learning (DDPG), and Reinforcement Learning from Imperfect Demonstrations (NAC), for Pommerman, predator-prey, and soccer as testbeds.

Publications

• Chopra, Tushar, and Vikash Kumar Sharma. “Droid Guard: An Approach to Make Android Secure.” International Journal of Computer Applications 117.8 (2015): 42-46.