How To Read This Directory

University robot learning labs are the source of most methods that later reach industrial arms and consumer robots. This directory lists 25 groups by focus area and principal investigator, drawn from each lab's own website. It is aimed at prospective PhD applicants, industry researchers scouting collaborations, and buyers trying to understand which academic lineage sits behind a product.

Focus areas are grouped into four buckets: manipulation and imitation learning, locomotion and legged control, medical and surgical robotics, and multi-agent / planning / HRI. Where a lab spans several, the primary bucket used by the lab itself is listed first. Named principal investigators are those the lab lists as director or head on its own site as of publication; check the lab page before contacting, because faculty rotate.

The Bot Scout does not rank these labs. Fit for a PhD, postdoc, or partnership depends on funding, cohort, and hardware access, none of which is visible from a public page. Use the directory to shortlist, then read the lab's recent publications and the individual advisor's student list.

The 25-Lab Directory

Each row cites the lab's own page. Where a lab operates under a broader institute (for example ETH RSL under D-MAVT, or JHU LCSR under the Whiting School), the institute link is used when the lab does not maintain a separate URL.

LabInstitutionPrimary focusPrincipal investigator(s)Source
Robotic AI & Learning Lab (RAIL)UC BerkeleyDeep RL, manipulation, foundation modelsSergey Levinerail.eecs.berkeley.edu
Malik Group (BAIR)UC BerkeleyVision-driven manipulation and locomotionJitendra Malikpeople.eecs.berkeley.edu/~malik
Personal Robotics LabUniversity of WashingtonAssistive manipulation, human-robot interactionSiddhartha Srinivasa (Dieter Fox leads UW Robotics & State Estimation Lab)personalrobotics.cs.washington.edu
Robotics & State Estimation LabUniversity of WashingtonPerception, manipulation, sim-to-realDieter Foxrse-lab.cs.washington.edu
Hamlyn Centre for Robotic SurgeryImperial College LondonSurgical robotics and medical imagingDaniel Elson (director)imperial.ac.uk/hamlyn-centre
Robotic Systems Lab (RSL)ETH ZurichLegged locomotion, field roboticsMarco Hutterrsl.ethz.ch
Computational Robotics Lab (CRL)ETH ZurichDesign, simulation, control co-optimizationStelian Coroscrl.ethz.ch
Multi-Scale Robotics Lab (medical)ETH ZurichMicrorobotics for medical applicationsBradley Nelsonmsrl.ethz.ch
IPRL / Bohg GroupStanford UniversityInteractive perception and manipulationJeannette Bohgweb.stanford.edu/group/iprl
ILIAD (Sadigh Group)Stanford UniversityInteractive and imitation learning, HRIDorsa Sadighiliad.stanford.edu
Movement Lab (Liu Group)Stanford UniversityCharacter animation and robot motor skillsC. Karen Liutml.stanford.edu
Robotics Institute (RI)Carnegie Mellon UniversityFull-stack robotics, manipulation, autonomyMatthew Johnson-Roberson (director)ri.cmu.edu
CSAIL RoboticsMITManipulation, planning, soft roboticsDaniela Rus (CSAIL director)csail.mit.edu/research/robotics
ARCLab (Su Group)UC San Diego3D perception and manipulationHao Sucseweb.ucsd.edu/~haosu
Intelligent Autonomous Systems (IAS)TU DarmstadtRobot learning, motor skillsJan Petersias.informatik.tu-darmstadt.de
High Performance Humanoid Technologies (H2T)Karlsruhe Institute of TechnologyHumanoid manipulation, ARMAR platformTamim Asfourh2t.iar.kit.edu
MIRMI (Munich Institute of Robotics and Machine Intelligence)TU MunichHuman-centered robotics, geriatronicsSami Haddadin (executive director)mirmi.tum.de
Autonomous Intelligent Systems LabUniversity of FreiburgLearning for navigation and manipulationWolfram Burgard, Abhinav Valadaais.informatik.uni-freiburg.de
Humanoid Robots Lab / Autonomous Intelligent SystemsUniversity of BonnHumanoid perception, cognitive roboticsSven Behnkeais.uni-bonn.de
Coordinated Science Laboratory / IRLUniversity of Illinois Urbana-ChampaignLearning-based control, planningKris Hauser and otherscsl.illinois.edu
Robotic Embedded Systems Laboratory (RESL)USCMulti-robot systems, learningGaurav Sukhatmerobotics.usc.edu/resl
Learning Agents Research Group (LARG)UT AustinReinforcement learning, multi-agentPeter Stonecs.utexas.edu/~larg
Institute for Robotics and Intelligent Machines (IRIM)Georgia TechFull-stack robotics, learning, HRISeth Hutchinson (executive director)research.gatech.edu/robotics
Neuroscience and Robotics Lab (NxR)Northwestern UniversityControl, learning, embodied intelligenceTodd Murpheymurpheylab.github.io
Laboratory for Computational Sensing + Robotics (LCSR)Johns Hopkins UniversityMedical robotics, CIS, image-guided surgeryRussell Taylor (founder, CIS)lcsr.jhu.edu

Corporate Research Groups Adjacent To Academia

Two industry groups publish at the same academic venues and collaborate with the labs above often enough to belong in a directory that a PhD applicant would use. Neither admits students, but both host interns and post code.

Google DeepMind's robotics team publishes under the DeepMind banner and has released the RT-1, RT-2, and RT-X model families and the Open X-Embodiment dataset with more than 20 academic partners. Details and citations are on the DeepMind publications page and the Open X-Embodiment site.

Nvidia's Seattle Robotics Lab, led by Dieter Fox (jointly with UW), releases Isaac Sim, Isaac Lab, and GR00T for humanoid foundation models; see developer.nvidia.com/isaac/lab. Toyota Research Institute's robotics program publishes on diffusion policy and large behavior models at tri.global/our-work/robotics.

  • Google DeepMind Robotics — RT-X, Open X-Embodiment, sim-to-real; internships only.
  • Nvidia Seattle Robotics Lab — Isaac Sim, Isaac Lab, GR00T; PhD internships with UW.
  • Toyota Research Institute (TRI) — diffusion policies, large behavior models; postdocs and interns.
  • Meta FAIR Embodied AI — Habitat simulator, Ego4D-linked robot learning.

What To Actually Look At Before Applying

A lab's reputation and its fit for a specific student are different questions. The public signals worth checking before an email or an application are: the advisor's recent first-author or last-author papers (not the lab's all-time list), current PhD-student output, hardware in the lab photos, and whether the lab releases code and datasets. All four are on the lab's own site.

Funding source is usually the least visible variable and the most important for a five-year PhD. NSF, DARPA, ONR, ERC, DFG, JST, and industry gifts each imply different publication cadence, IP rules, and travel. Ask directly.

Bottom Line

Use this directory as a shortlist, not a ranking. Read each lab's own page, look at the last two years of publications by the specific advisor you would work with, and confirm hardware and funding before you commit.

Bookmark the three labs whose recent work matches your target thesis, then email a specific PhD student in each before you write to the PI.

FAQs

Which university has the strongest robot learning program?

There is no single answer. Berkeley (RAIL, Malik), Stanford (Bohg, Sadigh, Liu), CMU RI, MIT CSAIL, ETH (RSL, CRL), and TU Munich (MIRMI) all consistently publish at CoRL, RSS, and ICRA and place graduates at top industry labs. Fit depends on subarea and advisor.

Which lab leads medical and surgical robotics?

Johns Hopkins LCSR founded the field of computer-integrated surgery under Russell Taylor, and Imperial's Hamlyn Centre and ETH's Multi-Scale Robotics Lab publish the most on image-guided and microrobotic surgery, respectively. All three list current projects on their sites.

Are corporate labs a substitute for a university PhD?

No. Google DeepMind Robotics, Nvidia Seattle Robotics, and Toyota Research Institute publish heavily but do not grant degrees. They do host PhD internships that often turn into full-time roles, and several PIs (Dieter Fox, Chelsea Finn, Karol Hausman) hold joint appointments.

How do I contact a principal investigator?

Read at least two recent papers first, then email a specific PhD student in the group with a concrete question about their work. If that goes well, write to the PI with a paragraph that names a project you would want to extend, not a generic statement of interest.

Do these labs release code and datasets?

Most do. Berkeley RAIL, Stanford IPRL and ILIAD, ETH RSL, TU Darmstadt IAS, and the DeepMind, Nvidia and TRI groups publish code and pretrained models with almost every paper. Check each lab's GitHub organization before applying.

Primary Sources