ROHAN THAKKER
  • Home
  • Research
  • Media
Picture

Rohan Thakker

I am a full-stack Roboticist. My research focus is on developing "tightly coupled" planning, control and vision algorithms to build resilient autonomy capabilies and deploy them on real-world missions.​ I have been working at NASA's Jet Propulsion Laboratory as a Robotics Research Technologist since 2017, prior to which, I got my graduate degree in Robotics at Carnegie Mellon Unversity, Pittsburgh, USA and undergrad in Mechanical Engineering at Visvesvaraya National Institute of Technology, Nagpur, India.
Google Scholar | LinkedIn | Twitter | Email | CV

Experience


Autonomy Lead
EELS, Snake Robot to Explore Enceladus

Sep 2022 – Present
​

Leading the Autonomy team of over 15 JPL researchers and engineers to develop risk-aware decision-making capabilities to achieve resilient mobility in perceptually degraded extreme environments.

Papers: <stay tuned>
Picture
Picture
​Learning for Planning Team Lead
High-speed Autonomous Off-road Navigation
DARPA RACER (Ex-1)

In collaboration with UC Berkely, MIT, and GaTech
Oct 2021 – Sep 2022
​

Led the collaboration with UC Berkeley and MIT to developed classical, imitation learning and reinforcement learning-based motion planning algorithms and deployed on full-scale off-road vehicles in Mojave Dessert.

Papers: [1], <stay tuned>
 
Guidance, Navigation and Controls Lead

DARPA Subterranean Challenge (Tunnel Circuit)
In collaboration with Caltech, MIT, Boston Dynamics, and more!
June 2018 – Feb 2020

Led the GNC group of Team Costar  to develop the full autonomous mobility stack for drones, legged robots, and wheeled robots deployed in real-worled environments such as the tunnels, mines, and caves. 
​Finished 2nd in Tunnel Circuit and 1st in Urban Circuit.

Papers: [1], [2], [3], [4], [5], [6], [7], [8]
Picture
Picture

Autonomy Researcher
Verifiable Mulit-agent Autonomy for Mars Helicopter and Rover 
In collaboration with Prof. Richard Murray & Prof. Aaron Ames, Caltech
​May 2017 – May 2018

Developed a framework for multi-agent risk-aware planning under uncertainty for coordination of Mars Rover and Mars Helicopter to generate policies that maximize the probability of satisfaction of a mission specification defined by Linear Temporal Logic (LTL) using POMDPs. 

Papers:
[1], [2], [3], [4], [5]

​Guidance, Navigation and Controls Developer
GPS-deined High-speed Autonomous Quadrotor Flight 

In collaboration with Google Tango Team, ARL
​Mar 2017 – Feb 2018
​

Developed planning and controls algorithms and software by exploiting geometric controllers and differential flatneess. Drone achieved speeds over 10m/s in indoor GPS-denied environments!
​
​
Papers: [1], [2], [3], [4]
Picture
Picture
Picture
Graduate Student Researcher, CMU
Kinodynamic Motion Planning for Dual Arm Manipulation & Non-linear control of quadrotors

Co-adivsed by Dr. Siddhartha Srinivasa and Dr. Koushil Sreenath
​
Developed a sampling-based Asymptotically Optimal Kinodynamic Motion Planner for high dimensional non-euclidean spaces. Demonstrated an order of magnitude improvement in convergence rate compared to state-of-the-art Hierarchical Rejection Sampling Algorithms.

Paper(s): [1]    Picture credits (1), (2)
  • Home
  • Research
  • Media