Dinesh Jayaraman (0000-0002-6888-3095)

Personal information

Biography

My research group, PennPAL (Perception, Action, and Learning), works at the intersections of robotics, computer vision, and machine learning. Our recent research directions center around ``efficient'' (experience- , supervision-, and compute-efficient) robot learning. Towards this, we are studying how notions of ``attention'' can be useful for robots during observation, representation, and policy learning. Another axis of our work investigates how layperson end users might easily specify tasks to a robot learner to teach it.

Activities

Collapse all

National Science Foundation: VA, VA, US

2023-02 | NSF CAREER Award

Distinction

Dinesh Jayaraman

https://orcid.org/0000-0002-6888-3095

Personal information

Emails & domains

Verified email domains

Websites & social links

Keywords

Countries

Biography

Activities

Professional activities (5)

National Science Foundation: VA, VA, US

Conference on Robot Learning: Auckland, NZ

Amazon: Seattle, US

Association for the Advancement of Artificial Intelligence: USA, US

IEEE Robotics and Automation Society: Piscataway, NJ, US

Works (50 of 83)

Articulate-Anything: Automatic Modeling of Articulated Objects via a Vision-Language Foundation Model

Learning to Achieve Goals with Belief State Transformers

REGENT: A Retrieval-Augmented Generalist Agent That Can Act In-Context in New Environments

The Value of Sensory Information to a Robot

TLControl: Trajectory and Language Control for Human Motion Synthesis

Vision Language Models are In-Context Value Learners

Can Transformers Capture Spatial Relations between Objects?

Composing Pre-Trained Object-Centric Representations for Robotics From “What” and “Where” Foundation Models

DrEureka: Language Model Guided Sim-To-Real Transfer

DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset

Environment Curriculum Generation via Large Language Models

Eureka: Human-Level Reward Design via Coding Large Language Models

Learning a Meta-Controller for Dynamic Grasping

Leveraging Symmetry to Accelerate Learning of Trajectory Tracking Controllers for Free-Flying Robotic Systems

Memory-Consistent Neural Networks for Imitation Learning

Open X-Embodiment: Robotic Learning Datasets and RT-X Models

Privileged Sensing Scaffolds Reinforcement Learning

Recasting Generic Pretrained Vision Transformers As Object-Centric Scene Encoders For Manipulation Policies

Task-Oriented Hierarchical Object Decomposition for Visuomotor Control

TLControl: Trajectory and Language Control for Human Motion Synthesis

Universal Visual Decomposer: Long-Horizon Manipulation Made Easy

ZeroFlow: Fast Zero Label Scene Flow via Distillation

ZeroMimic: Distilling Robotic Manipulation Skills from Web Videos

Training self-learning circuits for power-efficient solutions

Learning Policy-Aware Models for Model-Based Reinforcement Learning via Transition Occupancy Matching

LIV: Language-Image Representations and Rewards for Robotic Control

Planning Goals for Exploration

Prospective Learning: Principled Extrapolation to the Future

VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-Training

Vision-Based Contact Localization Without Touch or Force Sensing

Conservative and Adaptive Penalty for Model-Based Safe Reinforcement Learning

Discovering Deformable Keypoint Pyramids

Fighting Fire with Fire: Avoiding DNN Shortcuts through Priming

How Far I'll Go: Offline Goal-Conditioned Reinforcement Learning via $ f $-Advantage Regression

Know Thyself: Transferable Visuomotor Control Through Robot-Awareness

Long-HOT: A Modular Hierarchical Approach for Long-Horizon Object Transport

SMODICE: Versatile Offline Imitation Learning via State Occupancy Matching

Training Robots to Evaluate Robots: Example-Based Interactive Reward Functions for Policy Learning

An exploration of embodied visual exploration

Conservative Offline Distributional Reinforcement Learning

Embracing the Reconstruction Uncertainty in 3D Human Pose Estimation

Femtomolar SARS-CoV-2 Antigen Detection Using the Microbubbling Digital Assay with Smartphone Readout Enables Antigen Burden Quantitation and Dynamics Tracking

How Are Learned Perception-Based Controllers Impacted by the Limits of Robust Control?

Keyframe-focused visual imitation learning

Likelihood-Based Diverse Sampling for Trajectory Forecasting

Object Representations Guided By Optical Flow

SMiRL: Surprise Minimizing RL in Dynamic Environments

Cautious adaptation for reinforcement learning in safety-critical settings

Digit: A novel design for a low-cost compact high-resolution tactile sensor with application to in-hand manipulation

DIGIT: A novel design for a low-cost compact high-resolution tactile sensor with application to in-hand manipulation