I am broadly interested in the areas of Embodied Artificial Intelligence. My general interest lies in developing algorithmic formulations under real-world constraints, that are capable of synthesizing complex behaviors on physical hardware. More specifically, my current attention is focused on solving contact rich dexterous manipulation with free objects.
Biological designs are a result of years of evolution. They are effective, efficient and conceal within themselves enormous wisdom -- mechanical as well as intellectual. Complex movements exhibited by the animal kingdom is a result of a tightly integrated system of bio-muscular structures and fast reactive motor control. If artificial counterparts are to exhibit comparable behaviors, the solution is likely to be situated at the intersection of mechanism and computation. I use tools from reinforcement learning, robotics, optimal control, and optimization to reason at the intersection of mechanisms and algorithms to design robust sample-efficient techniques that scale to and succeed on the physical hardware.
(2022-2023) | |
---|---|
MyoDex: Generalizable Representations for Dexterous Physiological ManipulationVittorio Caggiano, Sudeep Dasari, Vikash KumarInternational Conference on Machine Learning (ICML) 2023 webpage | |
LIV: Language-Image Representations and Rewards for Robotic ControlJason Ma, Vikash Kumar, Amy Zhang, Osbert Bastani, Dinesh JayaramanInternational Conference on Machine Learning (ICML) 2023 webpage | |
SAR:Generalization of Dexterity via Synergistic Action RepresentationCameraon Berg, Vittorio Caggiano, Vikash KumarProceedings of Robotics: Science and Systems (RSS) 2023 webpage | |
ACT: Learning Fine-Grained Bimanual Manipulation with Low-Cost HardwareTony Zhao, Vikash Kumar, Sergey Levine, Chelsea FinnProceedings of Robotics: Science and Systems (RSS) 2023 webpage | |
Visual Dexterity: In-hand Dexterous Manipulation from DepthTao Chen, Megha Tippur, Siyang Wu, Vikash Kumar, Edward Adelson, Pulkit Agrawal(Under Review) webpage | |
GenAug: Retargeting behaviors to unseen situations via Generative AugmentationZoey Chen, Sho Kiami, Abhishek Gupta, Vikash KumarProceedings of Robotics: Science and Systems (RSS) 2023 webpage | |
H2R: Zero-Shot Robot Manipulation from Passive Human VideosHomanga Bharadhwaj, Abhinav Gupta, {Shubham Tulsiani*, Vikash Kumar*}under review webpage | |
Dexterous Manipulation from Images: Autonomous Real-World RL via Substep GuidanceKelvin Xu, Zheyuan Hu, Ria Doshi, Aaron Rovinsky, Vikash Kumar, Abhishek Gupta, Sergey LevineInternational Conference on Robotics and Automation (ICRA) 2023 webpage | |
MoDem: Accelerating Visual Model-Based Reinforcement Learning with DemonstrationsNicklas Hansen, Yixin Lin, Hao Su, Xiaolong Wang, Vikash Kumar, Aravind RajeswaranInternational Conference on Learning Representations (ICLR) 2023 webpage | |
CACTI: A Framework for Scalable Multi-Task Multi-Scene Visual Imitation LearningZhao Mandi, Homanga Bharadhwaj, Vincent Moens, Shuran Song, Aravind Rajeswaran, Vikash KumarWorkshop on Pre-training Robot Learning, CORL 2022 webpage | |
All the Feels: A dexterous hand with large area sensingRaunaq Bhirangi, Abigail DeFranco, Jacob Adkins, Carmel Majidi, Abhinav Gupta, Tess Hellebrekers, Vikash Kumarunder submission webpage | |
Real World Offline Reinforcement Learning with Realistic Data SourceGaoyue Zhou*, Liyiming Ke*, Siddhartha Srinivasa, Abhinav Gupta, Aravind Rajeswaran, Vikash KumarInternational Conference on Robotics and Automation (ICRA) 2023 webpage | |
VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-TrainingJason Ma, Shagun Sodhani, Dinesh Jayaraman, Osbert Bastani, {Vikash Kumar*, Amy Zhang*}International Conference on Learning Representations (ICLR) 2023 webpage | |
Learning Dexterous Manipulation from Exemplar Object Trajectories and Pre-GraspsSudeep Dasari, Abhinav Gupta, Vikash KumarInternational Conference on Robotics and Automation (ICRA) 2023 webpage | |
Translating Robot Skills: Learning Unsupervised Skill Correspondences Across RobotsT Shankar, Y Lin, A Rajeswaran, V Kumar, S Anderson, J OhInternational Conference on Machine Learning (ICML) 2022 webpage | |
Cross-Domain Transfer via Semantic Skill ImitationKarl Pertsch, Ruta Desai, Vikash Kumar, Franziska Meier, Joseph J. Lim, Dhruv Batra, Akshara RaiConference on Robot Learning (CoRL), 2022 webpage | |
Curiosity Driven Self-supervised Tactile Exploration of Unknown ObjectsJianren Wang*, Yujie Lu*, Vikash Kumar,webpage | |
R3M: A Universal Visual Representation for Robot ManipulationSuraj Nair, Aravind Rajeswaran, Vikash Kumar, Chelsea Finn, Abhinav GuptaConference on Robot Learning (CoRL) 2022 Scaling Robot Learning Workshop - ICRA 2022 | Best Paper Award webpage | |
Can Foundation Models Perform Zero-Shot Task Specification For Robot Manipulation?Yuchen Cui, Scott Niekum, Abhinav Gupta, Vikash Kumar and Aravind RajeswaranLearning for Decision and Control (L4DC) 2022 Scaling Robot Learning Workshop - RSS 2022 | Best Paper Award Finalist webpage | |
RRL: Resnet as representation for Reinforcement LearningRutav Shah, Vikash KumarInternational Conference on Machine Learning (ICML) 2021 webpage |
Benchmarks | |
---|---|
MyoSuite: A contact-rich simulation suite for musculoskeletal motor controlLearning for Decision and Control (L4DC) 2022MyoSim: A contact-rich simulation suite for musculoskeletal motor controlInternational Conference on Robotics and Automation (ICRA) 2022Vittorio Caggiano, Huawei Wang, Guillaume Durandau, Massimo Sartori, Vikash Kumar webpage | |
RB2: Robotic Manipulation Benchmarking with a TwistSudeep Dasari, Jianren Wang, Joyce Hong, Shikhar Bahl, Yixin Lin, Austin Wang, Abitha Thankaraj, Karanbir Chahal, Berk Calli, Saurabh Gupta, David Held, Lerrel Pinto, Deepak Pathak, Vikash Kumar, Abhinav GuptaNeurIPS, Datasets and Benchmarks Track, 2021 webpage | |
Benchmarking In-Hand ManipulationSilvia Cruciani, Balakumar Sundaralingam, Kaiyu Hang, Vikash Kumar, Tucker Hermans, Danica KragicIEEE Robotics and Automation Letters (RAL) 2020 Special Issue: Benchmarking Protocols for Robotic Manipulation webpage | |
ROBEL: RObotics BEnchmarks for LearningMichael Ahn, Henry Zhu, Kristian Hartikainen, Hugo Ponte, Abhishek Gupta, Sergey Levine, Vikash KumarConference on Robot Learning (CoRL) 2019 webpage | |
Multi-Goal Reinforcement Learning: Challenging Robotics Environments and Request for ResearchMatthias Plappert, Marcin Andrychowicz, Alex Ray, Bob McGrew, Bowen Baker, Glenn Powell, Jonas Schneider, Josh Tobin, Maciek Chociej, Peter Welinder, Vikash Kumar, Wojciech ZarembaTeach Report 2018 Paper, Movie, webpage, arxiv |
Behavior Synthesis | |
---|---|
Reset-Free Reinforcement Learning via Multi-Task Learning:
Abhishek Gupta*, Justin Yu*, Tony Z. Zhao*, Vikash Kumar*, Aaron Rovinsky, Kelvin Xu, Thomas Devlin, Sergey Levine
| |
A Game Theoretic Framework for Model-Based Reinforcement LearningAravind Rajeswaran, Igor Mordatch, Vikash KumarInternational Conference on Machine Learning (ICML) 2020 webpage | |
Emergent Real-World Robotic Skills via Unsupervised Reinforcement LearningArchit Sharma, Michael Ahn, Sergey Levine, Vikash Kumar, Karol Hausman, Shixiang GuRobotics: Science and Systems (RSS), 2020 webpage | |
Time Reversal as Self-SupervisionSuraj Nair, Mohammad Babaeizadeh, Chelsea Finn, Sergey Levine, Vikash KumarInternational Conference on Robotics and Automation (ICRA) 2020 webpage | |
The Ingredients of Real World Robotic Reinforcement LearningHenry Zhu, Justin Yu, Abhishek Gupta, Dhruv Shah, Kristian Hartikainen, Avi Singh, Vikash Kumar, Sergey LevineInternational Conference on Learning Representations (ICLR) 2020 webpage | |
Dynamics-Aware Unsupervised Discovery of SkillsArchit Sharma, Shixiang Gu, Sergey Levine, Vikash Kumar, Karol HausmanInternational Conference on Learning Representations (ICLR) 2020 webpage | |
Deep Dynamics Models for Learning Dexterous ManipulationAnusha Nagabandi, Kurt Konolige, Sergey Levine, Vikash KumarConference on Robot Learning (CoRL) 2019 webpage | |
Multi-Agent Manipulation via Locomotion using Hierarchical Sim2RealOfir Nachum, Michael Ahn, Hugo Ponte, Shane Gu, Vikash KumarConference on Robot Learning (CoRL) 2019 webpage | |
Relay Policy Learning: Solving Long-Horizon Tasks via Imitation and Reinforcement LearningAbhishek Gupta, Vikash Kumar, Corey Lynch, Sergey Levine, Karol HausmanConference on Robot Learning (CoRL) 2019 webpage, | |
Learning Latent Plans from PlayCorey Lynch, Mohi Khansari, Ted Xiao, Vikash Kumar, Jonathan Tompson, Sergey Levine, Pierre SermanetConference on Robot Learning (CoRL) 2019 webpage | |
Dexterous Manipulation with Deep Reinforcement Learning:Efficient, General, and Low-CostHenry Zhu*, Abhishek Gupta*, Aravind Rajeswaran, Sergey Levine, Vikash KumarInternational Conference on Robotics and Automation (ICRA) 2019 webpage | |
Learning Deep Visuomotor Policies for Dexterous Hand ManipulationDivye Jain, Andrew Li, Shivam Singhal, Aravind Rajeswaran, Vikash Kumar, Emanuel TodorovInternational Conference on Robotics and Automation (ICRA) 2019 webpage | |
Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and DemonstrationsDAPG: Demo Augmented Policy GradientVikash Kumar*, Aravind Rajeshwaran*, Abhiskek Gupta, John Schulman, Emanuel Todorov, and Sergey LevineProceedings of Robotics: Science and Systems (RSS) 2018 Paper, suppliment, Movie, webpage, arxiv | |
Divide-and-Conquer Reinforcement LearningDibya Ghosh, Avi Singh, Aravind Rajeswaran, Vikash Kumar, Sergey LevineInternational Conference on Learning Representations (ICLR) 2018 Paper, Movie, webpage, arxiv, OpenReview | |
Variance Reduction for Policy Gradient with Action-dependent BaselineCathy Wu, Aravind Rajeswaran, Yan Duan, Vikash Kumar, Alexandre M Bayen, Sham Kakade, Igor Mordatch, Pieter AbbeelInternational Conference on Learning Representations (ICLR) 2018 webpage | |
Learning Dexterous Manipulation Policies from Experience and ImitationVikash Kumar, Abhiskek Gupta, Emanuel Todorov, and Sergey LevineUnder Review Paper, Movie | |
BEST MANIPULATION PAPER AWARD
Optimal Control with Learned Local Models: Application to Dexterous ManipulationVikash Kumar, Emanuel Todorov, and Sergey LevineIEEE International Conference on Robotics and Automation (ICRA) 2016 Paper, Movie | |
Real-time behaviour synthesis for dynamic Hand-ManipulationVikash Kumar, Yuval Tassa, Tom Erez and Emanuel TodorovIEEE International Conference on Robotics and Automation (ICRA) 2014 Paper, Movie webpage | |
An integrated system for real-time Model Predictive Control of humanoid robotsTom Erez, Kendall Lowrey, Yuval Tassa, Vikash Kumar, Svetoslav Kolev, Emanuel TodorovHumanoids 2013 Paper, Overview, Dynamics look at Dynamic walking Dynamic tracking | |
Synthesis of Complex Behaviors with Optimal ControlTodorov Emanuel, Tassa Yuval, Erez Tom, Mordatch Igor, Kulchenko Paul, Kumar VikashComputational and Systems Neuroscience (COSYNE) 2013 Abstract, Movie |
Virtual reality | |
---|---|
MuJoCo HAPTIX: A Virtual Reality System for Hand ManipulationVikash Kumar and Emanuel TodorovIEEE-RAS International Conference on Humanoid Robots 2015 Paper, Movie |
Tracking & calibration | |
---|---|
Precise Calibration of Robots with small link lengths using Kinematic ExtensionsVisak Chadalavada, Vikash Kumar(Under progress) Paper, Movie | |
STAC: Simultaneous Tracking and CalibrationTingfan Wu, Yuval Tassa, Vikash Kumar, Javier Movellan, Emanuel TodorovHumanoids 2013 Paper, Movie |
Robot design | |
---|---|
A Low-cost and Modular, 20-DOF Anthropomorphic Robotic Hand: Design, Actuation and ModelingZhe Xu, Vikash Kumar, Emanuel TodorovHumanoids 2013 Paper, Movie webpage | |
Fast, strong and compliant pneumatic actuation for dexterous tendon-driven handsKumar Vikash, Todorov EmanuelIEEE International Conference on Robotics and Automation (ICRA) 2013 Paper, Movie webpage | |
Design of an anthropomorphic robotic finger system with biomimetic artificial jointsZhe Xhe, Kumar Vikash, Matsuoka Yoki, Todorov EmanuelIEEE International Conference on Biomedical Robotics and Biomechatronics (BioRob) 2012 Paper, Movie webpage | |
Self and Mutual learning in Robotic Arm, based on Cognitive systemsKumar Vikash, Patil Chetan, Sachan SachanInternational MultiConference of Engineers and Computer Scientists 2010 (Shortlisted for Best paper Award) Paper, Movie |
Manuscripts | |
---|---|
High performance pneumatics using Model Predictive ControlVikash Kumar, Visak CV, Emanuel TodorovPaper | |
Optimizing fuzzy multi-objective problems using fuzzy genetic algorithms and FZDT test functionsKumar Vikash, Chakroborty DebjaniPaper | |
The UW Hand: A Low-cost, 20-DOF Tendon-driven Hand with Fast and Compliant ActuationZhe X, Kumar V, Todorov E.Paper, Movie, |