Hello
I am Linfeng Zhao (赵林风), a Postdoctoral Scholar at Stanford University, working with Prof. Mykel Kochenderfer and Prof. Jeannette Bohg at the Stanford Robotics Center.
I finished Ph.D. (2025) at Khoury College of Computer Sciences of Northeastern University, advised by Prof. Lawson L.S. Wong, where I closely collaborated with MIT LIS group and Prof. Leslie Kaelbling, and Prof. Robin Walters. I interned at Meta, Boston Dynamics AI Institute, Amazon, and Microsoft Research Asia. Before, I worked with Prof. Hao Su at UC San Diego (2018-19).
My research focuses on building human-level general-purpose agents that can act in the physical world—robots that navigate homes, manipulate objects, and accomplish long-horizon tasks in open-world scenarios with unseen environments and goals. I develop abstractions for decision-making that decompose complex behaviors into compositional building blocks. I develop learning and planning approaches to enable agents to reason about the world and plan their actions at decision time for scalable, generalizable, and efficient decision-making systems.
PhD Thesis (2025)
Learning and Planning with Structured Abstraction for Embodied Decision-Making
PDFUpdates
2026/07
2026/05
Our GEM paper, Equivariant Open-vocabulary Pick and Place via Language Kernels and Patch-level Semantic Maps, was selected as a RA-L 2025 Best Paper Award Honorable Mention.
2025/10
Relocate to California for Postdoc research at Stanford University.
2025/09
Defended my PhD thesis, Learning and Planning with Structured Abstraction for Embodied Decision-Making (thesis PDF).
2025/08
Give invited talk at University of Washington (Dieter Fox's group).
2025/08
Give invited talk at Stanford University (Mykel Kochenderfer's SISL group).
2025/07
Give invited talk at University of Pennsylvania (Dinesh Jayaraman's group).
2025/06
Give invited talk at Tsinghua University (Yi Wu's group).
2025/06
Give invited talk at RSS 2025 Workshop: Sense-Plan-Act with Foundation Models (joint with Prof. Lawson Wong).
2025/05
Give invited talk at MIT (Luca Carlone's group).
2024/11
Visit Brown University to give an invited talk at Brown Robotics seminar on Learning for Planning and World Modeling, (slides, video coming).
2024/10
Attend IROS 2024 to present our paper on equivariant graph planning for navigation.
2024/10
Attend WAFR 2024 to present our paper on equivariant sampling-based control.
2024/08
A paper (equivariant sampling-based control) accepted at WAFR 2024.
2024/08
Attend RLC 2024 in UMass Amherst to present our paper on (map-guided navigation).
2024/08
Attend RSS 2024 to organize the workshop Geometric and Algebraic Structure in Robot Learning (Webpage, Recordings) and co-presented a paper on planning to practice on Spot.
2024/05
A paper (map-guided navigation) accepted at RLC 2024. This is based on workshop paper at NeurIPS 2020 Deep RL workshop.
2024/05
A paper from BDAI internship (planning to practice on Spot) accepted at RSS 2024.
2024/05
Start internship at Meta in Menlo Park this summer.
2024/04
Give a guest lecture at Northeastern RL class on Learning for Long-horizon Planning (slides).
2024/03
We are organizing a workshop at RSS 2024 on _Geometric and Algebraic Structure in Robot Learning_.
2024/01
A paper on equivariant graph planning accepted at RA-L with an option to present at IROS 2024.
2024/01
Give in-person talks at Tsinghua University IIIS (host: Huazhe Xu) & ME (host: Rui Chen) (slides, recording to come).
2023/11
Present a poster at Boston Symmetry Day at MIT on symmetry in control.
2023/06
Continuing summer internship at Boston Dynamics AI Institute.
2023/01
Starting internship at the new Boston Dynamics AI Institute locally in Boston.
2023/01
Two papers (symmetric planning, implicit differentiable planning) accepted at ICLR 2023.
2022/12
Physically attend NeurIPS in New Orleans.
2022/07
Give in-person oral talk at ICML 2022 on compositional generalization (slides) and poster presentation for equivariant world models.
2022/06
2022/06
Spotlight talk on symmetric planning at RSS 2022 workshop Implicit Representation for Robotic Manipulation.
2022/05
Oral talk on compositional generalization (virtually) at ICRA 2022 workshop Compositional Robotics.
2022/05
Our paper on compositional generalization is accepted to ICML 2022 as Long Presentation (117/5630, 2.1%, oral).
2022/05
Another paper on symmetric embedding for equivariant world models is accepted to ICML 2022 as Short Presentation (poster).
2022/03
Two papers (compositional generalization, symmetric planning) to appear at RLDM 2022 (non-archival).
2021/06
Summer internship at Amazon Science in Seattle.
Robot Demos
Research Focus
Abstraction and Representation
Learning structured representations (symmetry, compositionality) for efficient generalization.
Planning and World Modeling
Developing world models and planning algorithms for long-horizon reasoning.
Representative work
Decision-time Scaling
Scaling decision quality with more compute and interaction.
Representative work
Mobile Manipulation
Enabling diverse manipulation skills on mobile robot platforms.
Representative work
Selected Publications
Action Map Policy: Learning 3D Closed-loop Manipulation via Pixel Classification
Haojie Huang*, Zhang Ye, Linfeng Zhao, Boce Hu, Mingxi Jia, Yu Qi, Ahmed Agha, Dian Wang, Robert Platt†, Robin Walters†
arXiv 2026, In submission
TL;DR
Action Map Policy casts closed-loop 3D manipulation as dense pixel classification, preserving multimodal action distributions while predicting a full action chunk in one forward pass.
Pix2Act: Image-Space Manipulation Policies with Equivariant Augmentation
Haojie Huang*, Linfeng Zhao, Haotian Liu, Zhang Ye, Si-Yuan Huang, Mingxi Jia, Boce Hu, Fangzhou Lin, Yu Qi, Dian Wang, Robin Walters†, Robert Platt†
arXiv 2026, In submission
TL;DR
Pix2Act represents manipulation as continuous image-space keypoint trajectories, enabling equivariant augmentation and robust reconstruction of closed-loop 3D actions.
Retriever: Composing Closed-Loop Asynchronous Robot Programs
Linfeng Zhao, Haojie Huang, Jiayuan Mao, Weiyu Liu, Mykel Kochenderfer†, Lawson L.S. Wong†
arXiv 2026, In submission
TL;DR
Retriever is a programming model and runtime that we use to build hierarchical pipelines for long-horizon, partially observable robot manipulation.
Seeing is Believing: Planning to Perceive with Foundation Models and Act Under Uncertainty
Linfeng Zhao, Willie McClinton*, Aidan Curtis*, Nishanth Kumar, Tom Silver, Leslie Kaelbling, Lawson L.S. Wong
arXiv 2025
TL;DR
We use vision-language foundation models as uncertainty estimators inside a symbolic belief-space planner, enabling robots to plan to perceive and act under partial observability.
Learning and Planning with Structured Abstraction for Embodied Decision-Making
Linfeng Zhao
PhD Thesis (2025)
TL;DR
This thesis studies how structured abstractions, including symmetry, compositionality, and belief-state representations, can make learning and planning more efficient for embodied decision-making.
Practice Makes Perfect: Planning to Learn Skill Parameter Policies
Nishanth Kumar*, Tom Silver*, Willie McClinton, Linfeng Zhao, Stephen Proul, Tomás Lozano-Pérez, Leslie Kaelbling, Jennifer Barry
RSS 2024
TL;DR
We enable a robot to rapidly and autonomously specialize parameterized skills by planning to practice them. The robot decides what skills to practice and how to practice them. The robot is left alone for hours, repeatedly practicing and improving.
Equivariant Open-vocabulary Pick and Place via Language Kernels and Patch-level Semantic Maps
Mingxi Jia*, Haojie Huang*, Zhewen Zhang, Chenghao Wang, Linfeng Zhao, Dian Wang, Jason Xinyu Liu, Robin Walters, Robert Platt, Stefanie Tellex
IEEE RA-L 2025 (Best Paper Award Honorable Mention, top 5 / 1,700+ papers)
TL;DR
GEM combines patch-level semantic maps, language-conditioned visual relevancy, and equivariant policy learning for sample-efficient open-vocabulary pick-and-place manipulation.
E(2)-Equivariant Graph Planning for Navigation
Linfeng Zhao*, Hongyu Li*, Taskin Padir, Huaizu Jiang†, Lawson L.S. Wong†
IEEE RA-L 2024
IROS 2024 (Oral)
TL;DR
We study E(2) Euclidean equivariance in navigation on geometric graphs and develop message passing network to solve it.
Integrating Symmetry into Differentiable Planning with Steerable Convolutions
Linfeng Zhao, Xupeng Zhu*, Lingzhi Kong*, Robin Walters, Lawson L.S. Wong
ICLR 2023
RLDM 2022
TL;DR
We formulate how differentiable planning algorithms can exploit inherent symmetry in path planning problems, named SymPlan, and propose practical algorithms.
Scaling up and Stabilizing Differentiable Planning with Implicit Differentiation
Linfeng Zhao, Huazhe Xu, Lawson L.S. Wong
ICLR 2023
TL;DR
We differentiate through the Bellman fixed point to decouple forward planning from backward computation, enabling a flexible forward budget and constant backward cost in planning horizon.
Toward Compositional Generalization in Object‑Oriented World Modeling
Linfeng Zhao, Lingzhi Kong, Robin Walters, Lawson L.S. Wong
ICML 2022 (Long Presentation, top 2%)
RLDM 2022
TL;DR
We formulate compositional generalization in object-oriented world modeling, and propose a soft and efficient mechanism for practice.
Deep Imitation Learning for Bimanual Robotic Manipulation
Fan Xie*, Alexander Chowdhury*, M. Clara De Paolis Kaluza, Linfeng Zhao, Lawson L.S. Wong, Rose Yu
NeurIPS 2020
TL;DR
We present a deep imitation learning framework for robotic bimanual manipulation in a continuous state-action space.