Vikash Kumar: Research Scientist In Embodied AI, Hand ...

Trang chủ » Vikash Kumar » Vikash Kumar: Research Scientist In Embodied AI, Hand ...

Có thể bạn quan tâm

Vikash Kumar Twitter Twitter Github Google Scholar
  • Highlights
  • Impacts
  • Publications
  • Thesis
  • CV / Bio
  • About me

My research focuses on understanding the fundamentals of embodied movement across biological, digital, and electromechanical systems. I am interested in building artificial beings—both digital and physical—that are indistinguishable from humans in their appearance, spatial reasoning, and behavioral intelligence.

I leverage tools from Machine Learning, Biomechanics, Robotics, and Optimal Control to design sample-efficient algorithms that scale to deliver embodied abilities surpassing human-level dexterity and agility. I strongly believe its possible to realize artificial agengts indistinguishable from humans in our lifetime.

Recent news

  • [Nov 08th] Honored with Early Career Keynote at CoRL 2024, Munich
  • [Nov 04th] Invited talk at Next-Gen Robot Learning Symposium, TU Darmstadt
  • [Nov 05th] Invited talk at WCBM workshop
  • [Aug 18th] Honored with "Young Alumni Achiever's Award" from Indian Institute of Technology, Kharagpur

Video highlights

Human-Embodied Intelligence

MyoSuite - contact-rich framework for musculoskeletal motor control MyoDex: Representations for Dexterous Physiological Manipulation SAR: Generalization of Dexterity via Synergistic Action Representation MyoChallenge 2022: Learning contact-rich manipulation using a musculoskeletal hand

DexterousManipulation

Fine-Grained Bimanual Manipulation with Low-Cost Hardware Visual Dexterity: In-hand Dexterous Manipulation from Depth Projects/icons/avail.png Reset-Free Reinforcement Learning via Multi-Task Learning Planning with Deep Dynamics Models Adroit Manipulation Platform

Foundation Models for Robot Learning

R3M: A Universal Visual Representation for Robot Manipulation LIV: Language-Image Representations and Rewards for Robotic Control GenAug: Retargeting behaviors to unseen situations via Generative Augmentation VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-Training Can Foundation Models Perform Zero-Shot Task Specification For Robot Manipulation? RRL: Resnet as representation for Reinforcement Learning

Multi-Task / Hierarchical Learning

CACTI: A Framework for Scalable Multi-Task Multi-Scene Visual Imitation Learning Reset-Free Reinforcement Learning via Multi-Task Learning Relay Policy Learning Learning From Play Multi-Task Manipulation

Model basedlearning

Planning with Deep Dynamics Models Dynamics Aware Discovery of Skills Time Reversal as Self-Supervision Adroit Manipulation Platform picking up objects Adroit Manipulation Platform

Model-FreeLearning

Reset-Free Reinforcement Learning via Multi-Task Learning Demo Augmented Policy Gradients RL: Efficient, General & Low-Cost Ingredients of real world RL Robotics Benchmarks for Learning Manipulation via Locomotion

SupervisedLearning

Relay Policy Learning Learning From Play Demo Augmented Policy Gradients RL: Efficient, General & Low-Cost Manipulation via Experience&Imitation

ReinforcementLearning

Ingredients of real world RL Planning with Deep Dynamics Models Dynamics Aware Discovery of Skills Robotics Benchmarks for Learning Manipulation via Locomotion Adroit Manipulation Platform Divide-and-Conquer Reinforcement Learning

TrajectoryOptimization

Adroit Manipulation Platform1 Adroit Manipulation Platform2 Adroit Manipulation Platform2

Adroit Manipulation Platform

Adroit Manipulation Platform1

Physically realistic Virtual Reality

Adroit Manipulation Platform1 Adroit Manipulation Platform1 Adroit Manipulation Platform1

Darpa Robotics Challenge (DRC)

Adroit Manipulation Platform Adroit Manipulation Platform Adroit Manipulation Platform1 Adroit Manipulation Platform2

RobotDesign

A dexterous hand with large area sensing Robotics Benchmarks for Learning Adroit Manipulation Platform1 Adroit Manipulation Platform Adroit Manipulation Platform2 Adroit Manipulation Platform2

Benchmarking

MyoChallenge 2022: Learning contact-rich manipulation using a musculoskeletal hand RoboHive: A Unified Framework for Robot Learning Robotics Benchmarks for Learning Multi-Goal Reinforcement Learning: Challenging Robotics Environments and Request for Research Allegro Hand

Brain MachineInterface

Adroit Manipulation Platform1 Adroit Manipulation Platform Adroit Manipulation Platform Adroit Manipulation Platform1 Adroit Manipulation Platform2

Tracking andCalibration

Adroit Manipulation Platform1 Adroit Manipulation Platform

©2013 | Design by Vikash, CSE, University of Washington, Seattle, USA

Từ khóa » Vikash Kumar