Sashank Tirumala

I am a first year robotics graduate student at Carnegie Mellon University, where I work on fundamental problems on Robot Manipulation (generally from a learning perspective). I completed my undergrad in IIT Madras in Engineering Design.

Previously I have interned at Indian Institute of Sciences, Bangalore, where I worked on Imitation Learning, Reinforcement Learning and , Quadrupedal Locomotion under the guidance of Shishir Kolathaya and Bharadwaj Amrutur. I've received the Srikanth Sundarajan Award, Ms Latha & Sampath Srinath Prize , Mercedes Benz Drive Challenge Winner, Continental Fiction2Science Challenge Winner and the FIITJEE Scholarship Award.

Email  /  CV  /  Google Scholar  /  LinkedIn

profile photo
Research

I'm interested in realizing human-like dynamic motion for robots. Much of my research is about using machine learning, optimization and control algorithms to enable robots to reliably interact with the physical world.

Robust Quadrupedal Locomotion on Sloped Terrains: A Linear Policy Approach
Kartik Paigwar, Lokesh Krishna, Sashank Tirumala, Naman khetan, Aditya Sagi, Ashish Joglekar, Shalabh Bhatnagar, Ashitava Ghosal, Bharadwaj Amrutur, Shishir Kolathaya
4th Conference on Robot Learning (CoRL 2020), MIT, USA
arXiv / project page / github / video

Developed simple linear policies with reinforcement learning to realize rough terrain locomotion on quadruped robots.

Learning Stable Manoeuvres in Quadruped Robots from Expert Demonstrations
Sashank Tirumala, Sagar Gubbi, Kartik Paigwar, Aditya Sagi, Ashish Joglekar, Shalabh Bhatnagar, Ashitava Ghosal, Bharadwaj Amrutur, Shishir Kolathaya
29th International Conference on Robot and Human Interactive Communication (RoMan 2020), Naples, Italy
arXiv / project page / github / video

Developed a novel neural architecture that provided higher accuracy and required lesser data in order to demonstrate Complex Manoeuvres in quadruped robots.

Gait Library Synthesis for Quadruped Robots via Augmented Random Search
Sashank Tirumala, Aditya Sagi, Kartik Paigwar, Ashish Joglekar, Shalabh Bhatnagar, Ashitava Ghosal, Bharadwaj Amrutur, Shishir Kolathaya
arXiv, 2019
arXiv / github / video

Provides techniques to bridge the sim-to-real gap and quickly learn locomotive policies for quadruped robots.


Want to make such a website? source code,