Akshay Sarvesh
akshaysarvesh25@gmail.com
akshays25@tamu.edu

I am a Research Scientist at Meta working with Meta Reality Labs. I graduated with my PhD from Electrical and Computer Engineering at Texas A&M University. I was advised by: Swaminathan Gopalswamy and P. R. Kumar.

/ / /


Research Highlights

The goal of my research is to enable autonomous navigation of robots in unstructured terrains. My research interests lie in the domain of Motion Planning, Path Planning and Dynamic Control of robots(theoritical and field robotics). I use classical optimization techniques and meta reinforcement learning techniques to solve these problems. My research touches upon the following topics:
1. Meta Reinforcement-Learning
2. Adaptive Controls
3. Motion Planning, Path Planning
4. Dynamic Controls


Work Experience

1. ML Research Intern (Summer 2022) @ Meta Reality Labs; Worked on an exploratory project for designing an Contextual Recommendation System for Oculus VR

2. Engineer (2015-2017) @ Mercedes-Benz R&D; Worked on Collission avoidance systems for the autonomy stack


Teaching at Texas A&M

1. ECEN-214 - Spring 2023: Electric Circuits theory - Lecturer/Teaching Fellow

2. I have also been privileged to mentor 3 amazing batches of students in a Robotics Boot-camp. I led the creation, design and execution of this Robotics Boot-camp for Interns, Undergraduate and Graduate students who are interested in getting started with Robotics.

3. Teaching Assistant for ECEN-248 - Fall 2022: Digital Systems Design

4. Teaching Assistant for ECEN-214 - Spring 2022: Electric Circuits theory


Selected Publications and Projects

game

Meta Reinforcement-Learning based Contextual Adaptive-Control using Representations
Akshay Sarvesh , Bryan Yaggi, Dileep Kalathil, Srinivas Shakkottai, Swaminathan Gopalswamy
Submitted to ICRA, 2024 - Preprint coming soon

Reshaping Local Path Planner
Akshay Sarvesh , Austin Carroll, Swaminathan Gopalswamy
Published at RAL journal, Presented at ICRA'22 - Kyoto,Japan
game

Reshaping Visco-Elastic String based Path-Planner
Sarvesh Mayilvahanan , Akshay Sarvesh, Swaminathan Gopalswamy
Submitted to IROS, 2023

Reinforcement Learning with Sparse Rewards using Guidance from Offline Demonstration
Desik Rengarajan , Gargi Vaidya, Akshay Sarvesh, Dileep Kalathil, Srinivas Shakkottai
Published at ICLR - Spotlight paper, 2022

Autonomous Control of Robot-Vehicles by Estimating Friction characteristics of surface terrain

Developed a vision-based deep learning model that estimates the surface friction of various ground terrain such as grass, concrete, and dirt.

Integrated the perception algorithm with receding horizon MPC (Python, C++, ROS) to control mobile vehicles such as a Jeep Grand Cherokee.

Aerial-Ground Vehicle Co-ordination

Performed Research in Army Research Lab’s AFC-AGC (Autonomous Air-Ground Vehicle Co-ordination) Project.

Integrated ROS2 RTI-DDS secure Communication Paradigm with Silvus radios for robust, secure and long range network communication.

Wrote a ROS2 package that automatically allows selection of required ROS1 topics to be bridged.


Website desgin adopted from Georgia Gkioxari