Research Scientist · Google DeepMind

Takahiro
Miki

I work on Gemini Robotics. Before that, I taught legged robots to walk in the wild: mountains, caves, rubble, and places that look like other planets.

Projects Papers CV
Experimental · 3D

Expedition

Fly across one island that holds my field work: mountain trails, caves and mines, a volcano, a disaster site and a home. 33 stops, each with the real footage.

Enter the expedition →
Video: Google DeepMind, Gemini Robotics 2 (intelligent whole-body control)
Now · Google DeepMind

Gemini Robotics 2

Part of the Gemini Robotics team, building one model that sees, reasons and acts on any robot body. I focus on its intelligent whole-body control capability.

About this project →
Before DeepMind

Projects

Six years at ETH Zurich’s Robotic Systems Lab, teaching the ANYmal robot to see, walk and explore: work that appeared in Science Robotics four times.

All 10 projects →
2019–2025 · Robotic Systems Lab, ETH Zurich

Learning perceptive locomotion

Reinforcement learning that lets legged robots see the ground and still trust their legs.

Best Paper · ICRA 2024
2025–2026 · Robotic Systems Lab, ETH Zurich

Natural motion: animal priors and humanoids

Learning from recorded motion so robots move naturally, then adapting it to rough terrain.
2021–2023 · Open source

Elevation mapping and navigation

elevation_mapping_cupy: fast, GPU-based terrain maps for legged robots, and planning on top of them.
2021–2024 · Robotic Systems Lab, ETH Zurich

Legged manipulation

Legs that also handle things: opening doors, balancing an arm, catching a ball.
2019–2021 · Team CERBERUS

DARPA Subterranean Challenge

First place and the $2,000,000 prize in the 2021 Final.

1st place · DARPA SubT Final
2023–2026 · Field deployments

Science in the field

Taking legged robots to volcanoes, planetary analog sites and city streets.
Paper explainers

Every paper, explained in about 90 seconds

32 of my papers have an animated explainer: a short narrated video, the story in plain language, and the method step by step.

Browse the explainers →
Daily arXiv digest

Fresh from arXiv

Every day a bot picks a handful of new robotics and AI papers and writes a plain-language summary of each, in English and Japanese. These are the latest.

Open the digest →
10.01humanoid · locomotionLocomotion-Grounded Humanoid Soccer: Task-Gated Reinforcement Learning of a Multi-Directional Kicking Library 10.01humanoid · benchmarkFiatlux: A Long-Horizon Benchmark for Humanoid Ladder Climbing and Light-Bulb Replacement 10.01humanoid · controlCEER2: Directional and Tunable End-Effector and Root Compliance for Humanoid Loco-Manipulation 10.01quadruped · locomotionDense Temporal Motion Retargeting for Legged Robots 10.01humanoid · teleoperationNEXUS: Perceptive Whole-Body Control for Terrain-Adaptive Teleoperation 10.01drone · manipulationDrone Soccer: Learning to Manipulate with Multicopter Downwash

News

  1. Gemini Robotics 2 is out: our model that controls robots from tabletop arms to full humanoids.
  2. Started at Google DeepMind as a research scientist.
  3. Defended my PhD at ETH Zurich: “Bridging Perception and Control for Legged Locomotion and Navigation in the Wild”.
  4. Our paper received the IEEE ICRA Best Paper Award in Service Robotics.
  5. New co-authored paper published in Science Robotics.
  6. Our paper was selected as a finalist for the ICRA 2024 Best Paper Award in Service Robotics.
  7. Our video was selected for IEEE Spectrum’s Video Friday.
  8. Won 2nd prize in the SaTML LLM CTF competition with team RSLLM: $1,000 and a travel grant to the conference in Toronto.
  9. Uploaded “Learning to walk in confined spaces using 3D representation”.
  10. Two papers (one co-authored) accepted to ICRA 2024.
  11. Launched my homepage. It was more effort than I expected.