Artem Molchanov

Researcher in Robotics
and Machine Learning



About

I'm a researcher in robotics and machine learning. I have a position as a Senior Deep Learning Scientist in the Autonomous Vehicles division at NVidia. I completed my Ph.D. in the Robotics Embedded Systems Laboratory (RESL) at the University of Southern California under supervision of Gaurav Sukhatme.

My research interests are focused on systems that have to operate and adapt under sparse, noisy and incomplete information. Most relevant to robotics, this scenario requires careful leveraging of indirect sources of information to make intelligent decisions, such as priors acquired through simulation, additional sensors indirectly (nontrivially) measuring relevant quantities, communication with other agents, different indicators of task progression. Lately I have been working in the areas of reinforcement and imitation learning, where I am investigating learning under sparse rewards and a sample efficient adaptation to new tasks by combining simulation, automatic task curriculum and meta learning. I applied my research to a variety of platforms including quadrotors, robotic arms, underwater vehicles and autonomous cars. A more broad overview of my latest research direction can be found here.

During my Ph.D program at RESL I did a number of internships at:
Facebook AI Research (FAIR) (Summer, 2019) where I was part of the Robotics team and collaborated with Franziska Meier and Edward T. Grefenstette working on the project of meta-learning for RL;
Nvidia (Summer, 2017) where I was part of the Learning & Perception Research team lead by Jan Kautz and collaborated with Stan Birchfield and Jonathan Tremblay working on the project of curriculum learning for RL and the project on learning human-readable plans from demonstrations;
Volkswagen Electronics Research Lab (Summer, 2016) with the research team lead by Lutz Junge and collaborated with Premkumar Natarajan while working on neural network compression;
Blue River Technology (Summer, 2015) where I worked under supervision of Marci Meingast on improving a vision system of the lettuce thinning bot using deep learning.
Besides that, together with Vadim Butakov and Rob Simpson, I worked on a startup called Swerve.ai targeting improvement of safety in autonomous driving by enabling handling at the limits of friction.

I studied Mechanical and Electrical Engineering at Bauman Moscow State Technical University (BMSTU) and Computer Science at the University of Southern California (USC). In May 2010, I graduated cum laude obtaining Master of Engineering degree at BMSTU. In May 2015, I received Master of Computer Science degree from USC where I later defended my Ph.D in May 2020.

News

Research




Reinforcement & Imitation Learning

Decentralized Control of Quadrotor Swarms with End-to-end Deep Reinforcement Learning
International Conference on Robot Learning (CORL), 2021
S. Batra, Z. Huang, A. Petrenko, T. Kumar, A. Molchanov, G. Sukhatme
Meta-Learning via Learned Loss
International Conference on Pattern Recognition (ICPR), 2020
S. Bechtle, A. Molchanov, Y. Chebotar, E. Grefenstette, L. Righetti, G. Sukhatme, F. Meier
Sim-to-(Multi)-Real: Transfer of Low-Level Robust Control Policies to Multiple Quadrotors
International Conference on Intelligent Robots and Systems (IROS), 2019
A. Molchanov, T. Chen, W. Hönig, J. Preiss, N. Ayanian, G. Sukhatme
Region Growing Curriculum Generation for Reinforcement Learning
2018
A. Molchanov, K. Hausman, S. Birchfield, G. Sukhatme
Synthetically Trained Neural Networks
for Learning Human-Readable Plans from Real-World Demonstrations

International Conference on Robotics and Automation (ICRA), 2018
J. Tremblay, T. To, A. Molchanov, S. Tyree, J. Kautz, S. Birchfield

Perception for Manipulation

Contact Localization on Grasped Objects using Tactile Sensing
International Conference on Intelligent Robots and Systems (IROS), 2016
A. Molchanov, O. Kroemer , Z. Su, G. S. Sukhatme.
Force Estimation and Slip Detection for Grip Control using BioTac
International Conference on Humanoid Robotics (Humanoids), 2015
Z. Su, K. Hausman, Y. Chebotar, A. Molchanov, G. Loeb, G. Sukhatme, S. Schaal

Multi-Robot Systems Control

Active Drifters: Towards a Practical Multi-Robot System for Ocean Monitoring
International Conference on Robotics and Automation (ICRA), 2015
A. Molchanov, A. Breitenmoser, G. Sukhatme
Circling the Seas: Design of Lagrangian Drifters for Ocean Monitoring
IEEE Robotics & Automation Magazine (RAM), 2016
S. Subbaraya, A. Breitenmoser, A. Molchanov, Jorg Muller, Carl Oberg, D. Caron, G. Sukhatme

Workshop Publications

Meta Learning via Learned Loss
ICML Workshop on Multi-Task and Lifelong Reinforcement Learning, 2019
S. Bechtle, A. Molchanov, Y. Chebotar, E. Grefenstette, L. Righetti, G. Sukhatme, F. Meier. bibtex pdf

Sim-to-(Multi)-Real: Transfer of Low-Level Robust Control Policies to Multiple Quadrotors
Southern California Robotics Symposium, 2019
A. Molchanov, T. Chen, W. Honig, J. A. Preiss, N. Ayanian and G. S. Sukhatme. website bibtex pdf

Model-free Contact Localization for Manipulated Objects using Biomimetic Tactile Sensors
Humanoids Workshop on Tactile Sensing for Manipulation, 2016
A. Molchanov, O. Kroemer , Z. Su, G. S. Sukhatme. website bibtex pdf

BiGS: BioTac Grasp Stability Dataset
ICRA Workshop on Grasping and Manipulation Datasets, 2016
Y. Chebotar, K. Hausman, Z. Su, A. Molchanov, O. Kroemer, G. Sukhatme, S. Schaal
website bibtex pdf

Slip Classification using Tangential and Torsional Skin Distortions on BioTac
BMVA Workshop on Visual, Tactile and Force Sensing for Robot Manipulation, 2015
Z. Su, K. Hausman, Y. Chebotar, A. Molchanov, G. Loeb, G. Sukhatme, S. Schaal bibtex pdf

Slip Detection and Classification for Grip Control using Multiple Sensory Modalities on BioTac
IROS Workshop on Multimodal Sensor-Based Robot Control for HRI and Soft Manipulation, 2015
Z. Su, K. Hausman, Y. Chebotar, A. Molchanov, G. Loeb, G. Sukhatme, S. Schaal bibtex pdf

Active drifters: Sailing with the Ocean Currents
RSS Workshop on Autonomous Control, Adaptation, and Learning for Underwater Vehicles, 2014
A. Molchanov, A. Breitenmoser, G. S. Sukhatme bibtex pdf

Applied Research


Autonomous Driving

Low Friction Areas Localization Using Stereo Cameras and Microphones for Autonomous Driving
 Swerve.ai

A. Molchanov, V. Butakov, R. Simpson
Benchmarking Neural Network Compression Approaches in Application to Semantic Image Segmentation for Autonomous Driving
 Volkswagen Electronics Research Lab

A. Molchanov

Agricultural Robotics

Neural Network based Plant Detector and Classifier for Lettuce Thining Bot
 Blue River Technology

A. Molchanov, M. Meingast

Underwater Robotics

Control and State Estimation of a Remotely Operated Underwater Vehicle
 Research Institute of Special Mechanical Engineering: Underwater Vehicles Department

K. Chernenko, A. Molchanov, S. Egorov, A. Kutsenko