About me
I am a Ph.D. student at the École polytechnique fédérale de Lausanne (EPFL) and a research assistant at Robot Learning and Interaction group (RLI) under the supervision of Dr.Sylvain Calinon in the Idiap Research Institute. Recently, I finished a research internship at Disney Research Studios, Zurich. Prior to my PhD I have worked in the Engineering Development Group at The MathWorks Inc. and as Research Associate in the Robert Bosch Center for Cyber-Physical Systems, Bangalore.
My research in robotics lies at the intersection of data-efficient machine learning and control engineering. In particular, using tensor networks I am developing algorithms for Robot Control, Reinforcement Learning, and Motion Planning that were previously considered to be intractable.
If you are curious about my research, checkout my previous work which proposes a new methodology for global optimization of non-convex functions encountered in robotics. In my latest work, this is being used to learn some highly challenging control tasks in robotics, outperforming existing reinforcement learning algorithms for hybrid control.
What are tensor networks?
Tensor Networks (aka. low-rank tensor appoximation techniques) are powerful function approximation techniques heavily used in physics including quantum computing. They are particularly interesting for robotics as they offer interpretablity, algebraic manipulation and can model complex functions with high accuracy. See my publications for more details.
What can you do with tensor networks?
Motion Planning: Global Optimization + Multiple Solutions
Control of under-actuated systems:
Pendulum Swing-up
Cart-Pole Swing-up
Control of Systems with Hybrid Contact Dynamics:
Control of Systems with Hybrid State and Action Space:
Object Pushing Task
Real-world
Latest News
2024
- Our paper “Generalized Policy Iteration using Tensor Approximation for Hybrid Control” is accepted as a Spotlight paper (5% acceptance rate) at the International Conference on Learning Representations (ICLR 2024).
- Check how it provides a novel methodology for robot control Website
2023
- Our paper “Tensor Train for Global Optimization Problems in Robotics” is accepted for publication at The International Journal of Robotics Research (IJRR)
- Check how it bring new perspective into numerical optimization in robotics Website
[June - September] Summer internship at Disney Research Studios, Zurich. You can have a look here for more details about the project I was part of.
- Workshop paper “Dynamic Programming using Tensor Approximation for Contact-rich Manipulation” is presented in ICRA-2023
- workshop paper
- The main paper will be released soon (watchout for some cool results where we outperform RL algorithms on some highly challenging control tasks)
2022
- Our paper “Tensor Train for Global Optimization Problems in Robotics” is out on arXiv (under review at The International Journal of Robotics Research (IJRR))
2021
Our paper “Ergodic Exploration using Tensor Train: Applications in Insertion task” won the institute’s best paper award of the year (2021) from the Idiap Research Institute.
Our paper “Trajectory Prediction with Compressed 3D Environment Representation using Tensor Train Decomposition” is accepted at ICAR.
Our paper “Ergodic Exploration using Tensor Train: Applications in Insertion task” is accepted for publication at IEEE Transactions on Robotics.
2019
(June) Started my PhD at EPFL and as a Research Assistent at Idiap Research Institute, Martigny.
(Feb) Joined Robert Bosch Center for Cyber-Physical systems as a Research Associate
- I will be applying Reinforcement Learning to generate walking gaits for a quadruped called Stoc which is developed in-house.
2016
- Started working at The MathWorks Inc, India in the Engineering Development Group
- I will be working with MATLAB, Simulink, Robotics System Toolbox, Automated Driving System Toolbox.
2014
- Started a master’s degree in Mechanical Engineering at the Indian Institute of Science, Bangalore
- I will be focusing on Dynamic Systems and Control.