Embodied AI Researcher Tracker


ETH Zurich	Marco Hutter Scholar	Robotic Systems Lab (RSL) ETH Zurich RSL	Legged Locomotion Field Robotics Mobile Manipulation	ANYmal Unknown'26 ANYmal Parkour SciRob'24 ANYmal Parkour: Learning Agile Navigation Fully learned approach for quadrupedal robots navigating complex parkour obstacles using onboard perception. CERBERUS DARPA'21 ALMA IROS'19 SpaceHopper 2024 SpaceHopper: A Small-Scale Legged Robot for Low-Gravity Exploration A three-legged robot designed for hopping in low-gravity environments. Door Opening 2023 Learning to Open and Traverse Doors with a Legged Manipulator Autonomous door opening and traversal using a legged robot with an arm.
UC Berkeley	Sergey Levine Scholar	Robotic AI & Learning Lab (RAIL) UC Berkeley BAIR	Vision-Language-Action Robot Datasets Offline RL	Open X-Embodiment ICRA'24 Open X-Embodiment: Robotic Learning Datasets and RT-X Models Aggregates data from 22 robots across 21 institutions. RT-X demonstrates positive transfer across diverse robotic platforms. RT-1 RSS'23 RT-1: Robotics Transformer for Real-World Control at Scale Multi-task transformer trained on large-scale real-world data. Demonstrates zero-shot generalization to new tasks and environments. RT-2 CoRL'23 RT-2: Vision-Language-Action Models VLA model co-fine-tuned on robotic trajectory data and internet-scale tasks. Transfers web knowledge to robot control. CQL NeurIPS'20 D4RL arXiv'20 D4RL: Datasets for Deep Data-Driven Reinforcement Learning A benchmark for offline reinforcement learning. Bridge Data RSS'22 Bridge Data: A Large-Scale Dataset for Robot Learning A large-scale dataset for boosting generalization in robot learning.
Industry Research Labs	NVIDIA	GEAR Lab Generalist Embodied Agent Research	Simulation Reward Design Robot Foundation Models	GR00T 2024 Eureka ICLR'24 Eureka: Human-Level Reward Design via LLMs Uses LLMs to generate and refine reward functions for complex robotic skills through evolutionary optimization. Voyager arXiv'23 Isaac Lab Open Source Isaac Gym 2021 Isaac Gym NVIDIA's physics simulation environment for reinforcement learning. Key People: Jim Fan , Yuke Zhu
Stanford University	Chelsea Finn Scholar	IRIS Lab Stanford + Physical Intelligence	Bimanual Teleoperation Vision-Language-Action Meta-Learning	ALOHA RSS'23 ALOHA: A Low-cost Open-source Hardware System for Bimanual Teleoperation Low-cost bimanual teleoperation system for fine manipulation. Uses Action Chunking with Transformers (ACT) to learn complex tasks from few demonstrations. Mobile ALOHA CoRL'24 Mobile ALOHA: Learning Bimanual Mobile Manipulation Extends ALOHA with mobile base and whole-body teleoperation. Performs complex mobile manipulation tasks like cooking and cleaning through co-training. OpenVLA CoRL'24 OpenVLA: An Open-Source Vision-Language-Action Model 7B parameter open-source VLA model trained on 970K robot demonstrations. Generalist backbone for manipulation with efficient fine-tuning. MAML ICML'17 Model-Agnostic Meta-Learning Foundational meta-learning algorithm enabling fast adaptation to new tasks with minimal data. Meta-World CoRL'19 Meta-World A benchmark and evaluation for multi-task and meta-reinforcement learning.
Stanford University	Shuran Song Scholar	Robotics and Embodied AI Lab (REAL) Stanford REAL	Diffusion Policies Teleoperation Data Capture Contact-Rich Manipulation	Diffusion Policy RSS'23 Diffusion Policy: Visuomotor Policy Learning via Action Diffusion Represents robot policies as conditional denoising diffusion processes. Handles multimodal action distributions and high-dimensional action spaces with significant improvements over prior methods. UMI RSS'24 Universal Manipulation Interface A portable, hand-held data collection interface that allows diverse, in-the-wild data gathering for robot learning with zero-shot transfer to various robot arms. RoboPanoptes CoRL'24 RoboPanoptes Multi-camera system for robust robot perception and manipulation tasks. ACP CoRL'24 Adaptive Compliance Policy Learning adaptive compliance for contact-rich manipulation tasks. RoboCook ICRA'23 🏆 RoboCook: Long-Horizon Elasto-Plastic Object Manipulation Learning to manipulate elasto-plastic objects like dough.
Industry Research Labs	Google DeepMind (Gemini Robotics)	Gemini Robotics Team Google DeepMind	Vision-Language-Action Large-Scale Robotics World Models	RT-1 RSS'23 RT-2 CoRL'23 Genie ICML'24 Genie: Generative Interactive Environments 11B parameter foundation world model generating action-controllable environments from text/images without ground-truth labels. Dreamer V4 Unknown'26 Key People: Jie Tan , Raia Hadsell , Carolina Parada, Danijar Hafner
Industry Research Labs	Meta AI	FAIR Fundamental AI Research	World Models Self-Supervised Vision Embodied Simulation	V-JEPA 2024 V-JEPA: Video Joint-Embedding Predictive Architecture Learns visual representations by predicting masked spatio-temporal regions in latent space for improved world understanding. Habitat 3.0 Unknown'26 Habitat-Lab ICCV'19 Habitat-Lab A simulation platform for research in Embodied AI. V-JEPA 2 2025 V-JEPA 2: World Model Benchmarks Next-generation video joint-embedding predictive architecture. Former: Yann LeCun (Now AMI Labs)
Massachusetts Institute of Technology	Gabriel Margolis Scholar	Improbable AI Lab MIT CSAIL	Legged Locomotion Sim-to-Real Transfer Athletic Behaviors	Athletic Loco-Manipulation RSS'25 Bridging the Sim-to-Real Gap for Athletic Loco-Manipulation Introduces Unsupervised Actuator Net (UAN) for sim-to-real transfer of athletic whole-body behaviors. Enables robots to lift, throw, and drag with task rewards. Walk These Ways CoRL'23 Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior Fast, robust open-source locomotion controller with diverse gaits. Achieves rapid sim-to-real transfer through multiplicity of behavior training. Rapid Locomotion RSS'22 Rapid Locomotion via Reinforcement Learning End-to-end learned controller achieving record agility for MIT Mini Cheetah, sustaining speeds up to 3.9 m/s through RL. Extreme Parkour ICRA'24 Extreme Parkour with Legged Robots Enables quadruped robots to perform parkour over challenging obstacles using vision-based RL with single front-facing depth camera.
National University of Defense Technology	Hang Zhao Scholar	Embodied AI Lab Wuhan University + NUDT	World Models Industrial Robotics Combinatorial Optimization	PIN-WM RSS'25 PIN-WM: Learning Physics-INformed World Models for Non-Prehensile Manipulation Physics-informed world model using differentiable simulation and Gaussian Splatting for non-prehensile manipulation. Enables robust sim-to-real transfer with physics-aware digital cousins. 3D Bin Packing (IJRR) IJRR'25 Deliberate Planning of 3D Bin Packing on Packing Configuration Trees Industrial packing robot system with deliberate planning on packing configuration trees. Successfully deployed at Deli Smart Logistics Park. Physically Realizable Packing SIGGRAPH'24 Learning Physically Realizable Skills for Online Packing of General 3D Shapes Learning-based approach for online packing of irregular 3D shapes with physical constraints. Combines RL with differentiable physics simulation. PCT ICLR'22 Learning Efficient Online 3D Bin Packing on Packing Configuration Trees First learning-based method for online 3D bin packing using tree-structured state representation. Enables efficient policy learning for combinatorial optimization.
Stanford University	Fei-Fei Li Scholar	Stanford Vision and Learning Lab (SVL) Stanford SVL + World Labs	Household Simulation Activity Benchmarks World Models	iGibson CoRL'21 iGibson 2.0: Object-Centric Simulation Open-source simulation supporting object states (temperature, wetness) for household robot learning with VR demonstration collection. OmniGibson CoRL'22 OmniGibson: A Platform for Embodied AI Photorealistic simulation with physics for BEHAVIOR-1K benchmark. Enables long-horizon human-centered robotic tasks. BEHAVIOR-1K CoRL'22 BEHAVIOR-1K: A Human-Centered Benchmark 1,000 everyday activities across 50 scenes with 5,000+ objects for comprehensive embodied AI evaluation. World Labs Startup World Labs Spatial intelligence company building Large World Models (LWMs) for 3D world perception and generation.
UC Berkeley	Pieter Abbeel Scholar	Berkeley Robot Learning Lab UC Berkeley BAIR	World Models Model-Based RL Decision Transformers	DayDreamer CoRL'22 DayDreamer: World Models for Physical Robot Learning Learns world models directly on physical robots without simulators. Masters walking and manipulation in record time. DreamerV3 ICLR'23 Mastering Diverse Domains through World Models Scalable world-model RL with fixed hyperparameters. Collects diamonds in Minecraft from scratch. Decision Transformer NeurIPS'21 SAC ICML'18
Carnegie Mellon University	Guanya Shi Scholar	Learning and Control for Agile Robotics (LeCAR) CMU LeCAR Lab	Humanoid Retargeting Dexterous Hands Agile Control	OmniRetarget arXiv'24 OmniRetarget: Interaction-Preserving Data Generation for Humanoid An interaction-preserving data generation engine based on an interaction mesh that explicitly models the geometric relationship between the robot and its environment. SPIDER arXiv'24 SPIDER: Scalable Physics-Informed Dexterous Retargeting A dynamically feasible, cross-embodiment retargeting framework for both humanoid and dexterous hand. Human -> physics -> real at scale. BFM-Zero arXiv'24 BFM-Zero: A Promptable Behavioral Foundation Model for Humanoid Control Using Unsupervised RL BFM-Zero enables zero-shot goal reaching, tracking, and reward optimization (any reward at test time) from one policy.
Massachusetts Institute of Technology	Pulkit Agrawal Scholar	Improbable AI Lab MIT CSAIL	Dexterous Manipulation Hierarchical Planning Skill Learning	DEXOP arXiv'25 HiP NeurIPS'23 Hierarchical Planning with Foundation Models Uses three foundation models (language, vision, action) for long-horizon robot planning. Vegetable Peeling CoRL'23
Peking University	Hao Dong Scholar	PKU-Agibot Lab PKU Hyperplane	Embodied Navigation Instruction Following Alignment	TensorLayer Open Source InstructNav arXiv TwinAligner arXiv
University of Texas at Austin	Yuke Zhu Scholar	Robot Perception and Learning Lab UT Austin RPL + NVIDIA	Robot Learning Manipulation Continual Learning	ASAP RSS'25 ASAP: Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills Two-stage framework for humanoid whole-body control that uses delta action learning to bridge sim-to-real dynamics mismatch. Enables highly agile motions on Unitree G1. GROOT CoRL'23 Learning Generalizable Manipulation Policies with Object-Centric 3D Representations Imitation learning method with object-centric and 3D priors for robust manipulation policies that generalize across backgrounds, viewpoints, and object instances. LOTUS ICRA'24 LOTUS: Continual Imitation Learning for Robot Manipulation Continual learning algorithm that enables robots to continuously learn new manipulation tasks while retaining previous skills through meta-controller composition.
Carnegie Mellon University	Abhinav Gupta Scholar	Abhinav Gupta's Lab CMU RI (President @ Skild AI)	Curiosity-Driven RL Self-Supervised Learning Robot Foundation Models	Skild Brain Startup Curiosity Learning ICML'17
Carnegie Mellon University	Deepak Pathak Scholar	Deepak Pathak's Group CMU RI (CEO @ Skild AI)	Curiosity-Driven RL Self-Supervised Exploration Robot Foundation Models	Curiosity-driven RL ICML'19 Skild Brain Startup
ETH Zurich	Davide Scaramuzza Scholar	Robotics and Perception Group (RPG) UZH / ETH Zurich	Agile Drone Flight Autonomous Racing	Agile Drone Flight Unknown'26 Autonomous Drone Racing Nature'23
Industry Research Labs	Physical Intelligence	Robot Foundation Models Physical Intelligence	Robot Foundation Models Vision-Language-Action Flow Models	pi0 2024 pi0 (π0): General-Purpose Robot Foundation Model Trained on diverse multi-task/multi-robot data. Combines VLMs with flow-matching for versatile robot control. pi0 Paper arXiv Founders: Karol Hausman (CEO), Sergey Levine , Chelsea Finn
Industry Research Labs	Vincent Vanhoucke Scholar	Waymo Research Waymo / Google	Embodied LLMs Task Planning Vision-Language-Action	PaLM-E ICML'23 PaLM-E: An Embodied Multimodal Language Model A large-scale embodied multimodal language model for robot control. SayCan CoRL'22 SayCan: Extracting Algorithmic Knowledge from LLMs Using LLMs to plan and execute tasks in the real world.
Massachusetts Institute of Technology	Leslie Kaelbling Scholar	Learning and Intelligent Systems (LIS) MIT CSAIL	Task and Motion Planning Symbolic Planning	TAMP arXiv'20 Embodied Intelligence Unknown'26
Peking University	He Wang Scholar	Embodied Perception and InteraCtion (EPIC) PKU EPIC Lab	6D Pose Estimation Category-Level Manipulation	NOCS CVPR'19 Category-Level Manipulation Unknown'26
Peking University	Yixin Zhu Scholar	CoRe Lab PKU CoRe Lab	Value Alignment Dexterous Hands Bimanual Manipulation	Tong Science Robotics Tong: A General Artificial Intelligence Agent A general AI agent capable of bidirectional human-robot value alignment. Bi-DexHands ICML'22 Bi-DexHands Bimanual Dexterous Manipulation with Deep Reinforcement Learning.
Shanghai Jiao Tong University	Cewu Lu Scholar	Machine Vision and Intelligence Group (MVIG) SJTU MVIG	Pose Estimation Robot Vision	AlphaPose ICCV'17 RoboTube Unknown'26
Stanford University	Jiajun Wu Scholar	Computational Vision and Geometry Lab Stanford SVL	Simulation Platforms Household Activities Benchmark Design	iGibson 2.0 CoRL'21 BEHAVIOR CoRL'22
Stanford University	Karen Liu Scholar	The Movement Lab Stanford TML	Differentiable Physics Physics-Based Animation Dynamics Modeling	Nimble Physics NeurIPS'21 Physics-based Animation Unknown'26
Stanford University	Silvio Savarese Scholar	Stanford / Salesforce Research Stanford Adjunct / Salesforce Chief Scientist	Learning from Demonstration Manipulation Datasets	RoboMimic CoRL'21 RoboMimic: A Modular Framework for Robot Learning from Demonstration A framework for robot learning from demonstration with standardized datasets. RoboMimic CoRL'21 RoboMimic A modular framework for robot learning from demonstration.
Tsinghua University	Huazhe Xu Scholar	Tsinghua Embodied AI Lab (TEA Lab) Tsinghua TEA Lab	Reward Learning Simulation Benchmarks Skill Learning	Diff-Reward arXiv ManiSkill Unknown'26
UC Berkeley	Jitendra Malik Scholar	Malik Lab UC Berkeley BAIR	Instance Segmentation Scene Understanding Embodied Simulation	Mask R-CNN ICCV'17 Habitat Unknown'26
UC Berkeley	Ken Goldberg Scholar	Berkeley AUTOLab UC Berkeley BAIR	Dexterous Grasping Cloud Robotics	Dex-Net RSS'17 FogROS ICRA'21
California Institute of Technology	Aaron Ames Scholar	AMBER Lab Caltech AMBER Lab	Bipedal Locomotion Multi-Robot Teams Humanoid Control	X1 Multi-Robot 2025 X1: First multi-robot team with humanoid and drone A multi-robot team combining humanoid robots and drones.
California Institute of Technology	Sangbae Kim Scholar	Biomimetic Robotics Lab MIT Biomimetic Robotics	Humanoid Robotics Legged Locomotion Robotics Hardware	MIT Humanoid 2024 Design and Development of the MIT Humanoid A dynamic humanoid robot capable of human-level athletic feats.
California Institute of Technology	Tomás Lozano-Pérez Scholar	Learning and Intelligent Systems (LIS) MIT CSAIL	Task and Motion Planning Manipulation Planning	Task & Motion Planning Unknown'26 Task and Motion Planning (TAMP) Foundational work on integrating high-level task planning with low-level motion planning.
Carnegie Mellon University	David Held Scholar	Robots Perceiving and Doing (RPAD) CMU Robotics Institute	Deformable Manipulation Manipulation Learning	Deformable Manipulation Unknown'26
Industry Research Labs	AMI Labs	AMI Labs AMI Labs	World Models Predictive Modeling	World Models 2026 Next-Gen Superintelligent AI Developing world models that understand physics and the physical world. Key People: Yann LeCun , Alexandre LeBrun
Industry Research Labs	AMI Labs	World Models / JEPA AMI Labs	JEPA Representation Learning World Models	JEPA Architecture Unknown'26 Founder: Yann LeCun (Executive Chairman)
Industry Research Labs	Google DeepMind (Open-Endedness)	Open-Endedness / Hardware Google DeepMind	World Models Generative Environments	Genie 2 2024 Genie 2: A Large-Scale Foundation World Model A foundation world model generating action-controllable environments. Key People: Aaron Saunders , Tim Rocktäschel , Jake Bruce , Michael Dennis
Industry Research Labs	Physical Intelligence	π Lab Physical Intelligence	Robot Foundation Models Vision-Language-Action Flow Models	π0 2024 π0: A Vision-Language-Action Flow Model A general-purpose foundation model for robot control. Key People: Lachy Groom , Brian Ichter , Sergey Levine
Industry Research Labs	Scale AI	Scale AI Scale AI	Data Labeling Model Evaluation	Data Labeling Unknown'26 Scale AI: Accelerating AI Development Providing high-quality data for training and evaluating large language models and autonomous systems. Key People: Alexandr Wang
Industry Research Labs	Skild AI	Shared Robot Brain Skild AI	Robot Foundation Models Robotic Platforms	Skild Brain 2024 Founders: Deepak Pathak (CEO), Abhinav Gupta (President)
Industry Research Labs	World Labs	Spatial Intelligence Large World Models	World Models Spatial Intelligence 3D Understanding	Large World Models Unknown'26 Founder: Fei-Fei Li
Massachusetts Institute of Technology	Russ Tedrake Scholar	Robot Locomotion Group MIT CSAIL	Multibody Dynamics Optimization Control Systems	Drake Open Source
National University of Defense Technology	Kai Xu Scholar	Embodied AI Lab NUDT	World Models Physics Simulation 3D Vision	PIN-WM RSS'25 PIN-WM: Learning Physics-INformed World Models for Non-Prehensile Manipulation Physics-informed world model using differentiable simulation and Gaussian Splatting for non-prehensile manipulation. Enables robust sim-to-real transfer with physics-aware digital cousins.
Peking University	Jingtao Ding Scholar	Autonomous Systems PKU / Tsinghua	World Models Model-Based RL Survey	World Models ACM'25 A Comprehensive Survey of World Models Systematic review of world models for understanding and predicting the future.
Peking University	Liwei Wang Scholar	Machine Learning Group PKU ML	Learning Theory Generalization Optimization	ML Foundations Unknown'26 Foundations of Machine Learning Theoretical research on the strengths and weaknesses of learning algorithms.
Peking University	Yang Liu Scholar	Wangxuan Institute PKU CS	Task-Oriented QA Visual Reasoning Human-Robot Interaction	Robot QA Unknown'26 Task-Driven Question Answering for Robots Research on interactive AI and visual reasoning for in-home robots.
Peking University	Zhibo Pang Scholar	School of Advanced Manufacturing and Robotics PKU AMR	System Orchestration Embodied Context Robotics Deployment	Embodied Context Protocol Unknown'25 Orchestrating Embodied Systems through the Embodied Context Protocol A framework for system deployment and coordination in embodied intelligence.
Shanghai Jiao Tong University	Hesheng Wang Scholar	Intelligent Robot & Machine Vision Lab (IRMV) SJTU IRMV (IROS 2025 Chair)	Visual Servoing Robotic Control	Visual Servoing Unknown'26
Shanghai Jiao Tong University	Weidong Chen Scholar	Autonomous Robot Laboratory SJTU Robotics	Medical Robotics Surgical Systems	Medical Robotics Unknown'26 Intelligent Medical Robotics Research on autonomous systems for medical and surgical applications.
Shanghai Jiao Tong University	Yong-Lu Li Scholar	RHOS Lab SJTU RHOS Lab	Human Activity Knowledge Vision-Language Reasoning	HAKE Unknown'26 Human Activity Knowledge Engine A knowledge-driven system for human activity understanding and reasoning.
Technical University of Munich	Alin Albu-Schäffer Scholar	Institute of Robotics and Mechatronics TUM / DLR Robotics	Humanoid Robotics Robot Mechatronics	Next-Level Robotics 2025 Next-Level Robotics: Humanoid Robotics as a Key Technology of Tomorrow Global developments in humanoid robotics.
Tsinghua University	Jianyu Chen Scholar	Jianyu Chen's Group Tsinghua IIIS	Humanoid Locomotion World Model Learning Sim-to-Real	DWL RSS'24
Tsinghua University	Yi Wu Scholar	IIIS Tsinghua IIIS	Multi-Agent RL Strategic Agents	MARL Unknown'26 Multi-Agent Reinforcement Learning Research on reinforcement learning for strategic language agents and human-robot interaction.
UC Berkeley	Koushil Sreenath Scholar	Hybrid Robotics Lab Berkeley Hybrid Robotics	Humanoid Control Legged Locomotion	Berkeley Humanoid arXiv'24 Berkeley Humanoid: A Research Platform for Learning-based Control A reliable and low-cost mid-scale humanoid research platform.