Naoki Shitanda

Profile photo

News

  • 2026-04-01: I received my M.S. degree and entered the Ph.D. program with the BOOST NAIS fellowship.
  • 2026-02-20: This webpage is now live.
  • 2026-01-26: First-author paper accepted to ICLR 2026.

Contact

shitanda[at]mi.t.u-tokyo.ac.jp

Google Scholar

LinkedIn

GitHub

Hi! I am a Ph.D. student in the Graduate School of Information Science and Technology at The University of Tokyo. I am affiliated with the Machine Intelligence Lab and the RIKEN AIP Robot Learning Team, and am advised by Prof. Tatsuya Harada and Dr. Takayuki Osa.

I am particularly interested in scalable reinforcement learning algorithms for large-scale parallel environments and in dexterous manipulation. Related work was accepted to ICLR 2026.

In summer 2025, I had a great opportunity to work on a dexterous manipulation project Soft Robotics Lab at ETH Zurich. I received my M.S. from The University of Tokyo and my B.Eng. from Kyoto University. Since 2021, I have also been working as a Research Assistant at Sony CSL - Kyoto under Prof. Jun Rekimoto.

🎉 Latest Paper 🎉

Rethinking Policy Diversity in Ensemble Policy Gradient in Large-Scale Reinforcement Learning

Paper image

ICLR 2026 To Appear (Poster)

We study multi-policy reinforcement learning in large-scale parallel environments and show how policy diversity affects learning. By appropriately controlling diversity within policy sets, our method achieves high sample efficiency and strong final performance on object manipulation tasks.

Project Page Read More

Education

  • 2026.04 - Present: Ph.D. Program, Department of Mechano-Informatics, Graduate School of Information Science and Technology, The University of Tokyo Supervisor: Prof. Tatsuya Harada
  • 2024.04 - 2026.03: Master of Information Science and Technology, Department of Mechano-Informatics, Graduate School of Information Science and Technology, The University of Tokyo Supervisors: Prof. Tatsuya Harada, Assoc. Prof. Takayuki Osa
  • 2019.04 - 2024.03: B.E., Department of Electrical and Electronic Engineering, Kyoto University Supervisor: Prof. Toshiya Doi

Study Abroad

  • 2024.07 - 2026.09: ETH Zurich ETH Robotics Summer Fellowship 2025, Prof. Robert Katzschmann, acceptance rate: 3.1%
  • 2022.09 - 2023.06: University of Bristol / Bristol Robotics Laboratory Exchange student & visiting researcher, Prof. Jonathan Rossiter

Internships

  • 2025.11 - Present: RIKEN Center for Advanced Intelligence Project (AIP)
    Robot Learning Team (Takayuki Osa)
  • 2021.02 - Present: Sony Computer Science Laboratories, Inc.
    Research assistant under Deputy Director Jun Rekimoto (Kyoto): (1) data collection support systems for 3D reconstruction (2025-), (2) in-the-wild 3D reconstruction (2023-2024), (3) interactive architectural elements with variable transparency (2022-2023)
  • 2024.07 - 2025.11: FingerVision Inc. Contract development of a computer vision system for automatic dish identification and dishwashing robots
  • 2024.05 - 2024.06: Preferred Networks, Inc.
    Part-time engineer (robot learning); contract ended after two months due to business-side decision.
  • 2019.10 - 2021.09: Muscle Corporation
    Part-time engineer; developed a low-cost mannequin robot for fashion displays. Adopted in the "Hiroko Koshino Exhibition" at Hyogo Prefectural Museum of Art from April 2022.

Other

  • 2025.06 - 2026.03: MITOU Program for Innovative IT Human Resources (FY2025)
    Project: development platform for modular social robots for self-expression, with Sotaro Yokoi, Naoki Shitanda, and Kento Matsuo. Acceptance rate: 9.3%

Peer-reviewed International Conference Papers

  1. Rethinking Policy Diversity in Ensemble Policy Gradient in Large-Scale Reinforcement Learning
    Naoki Shitanda, Motoki Omura, Tatsuya Harada, Takayuki Osa
    International Conference on Learning Representations 2026 (ICLR)
    To Appear (Poster)
    [Read More]
  2. Gaussians in the City: Enhancing 3D Scene Reconstruction under distractors with Text-guided Segmentation and Inpainting
    Naoki Shitanda, Jun Rekimoto
    SIGGRAPH Asia 2024 in Tokyo (Poster)
    Dec 2024, From Sony CSL Kyoto
    [Read More]

Peer-reviewed International Workshops

  1. Vision and Tactile Pose Identification for Picking a Target without Collision
    Saekwang Nam, Loong Yi Lee, Max Yang, Naoki Shitanda, Jonathan Rossiter, Nathan Lepora
    ViTac Workshop at IEEE International Conference on Robotics and Automation 2023 in London [ICRA2023 Workshop]
    May 2023
    [Read More]

Domestic Conferences

  1. Squama Yukimi: Interactive Architectural Elements in a Tea Room
    Naoki Shitanda, Jun Rekimoto
    AIJISA 2022 Interactive Session
    December 2022
    [Read More]

Competitions

  1. Team FANTASTIC
    Saekwang Nam, Loong Yi Li, Max Yang, Lihaoya Tan, Naoki Shitanda
    RoboSoft 2023 @Singapore Food Manipulation Task
    March 2023
    Competed in the food manipulation task with a team of two postdocs and two PhD students. I was responsible for the computer vision part.

Research Funding

  • GCL FY2025 Independent Research and Activity Grant, Category B FY2025, JPY 500,000

Scholarships

  • BOOST NAIS: Next-generation AI for Intelligent Society 2026 April ~ 2029 March, Fellowship + Research Funding(500,000 JPY/year)
  • Teijin Scholarship Foundation 2026 April ~ 2029 March
  • WINGS-IIW, The University of Tokyo (International Graduate Program for Intelligent Society) 2024 April ~ 2026 March (2029 March), Fellowship
  • Hayashi Rheology Memorial Foundation 2024 April ~ 2026 March (Completed)
  • Okuyama Scholarship Foundation 2022 September ~ 2024 March (Completed)
  • Leave a Nest Scholarship, Osaka Seminar FY2023 (Completed)
  • JASSO Overseas Study Support Program (Exchange Dispatch) + Special Continuation Scholarship 2022 September ~ 2023 May
  • Kyoto University International Education Program 2022 ~ 2024 (Completed)
  • Kyoto University Study Abroad Grant 2022 (Declined / Returned)
  • Okura Mutual Aid Association: Support for Students Studying Abroad 2022 (Completed)
  • KEYENCE Foundation "Go for it! Japanese University Students" Support Grant 2022 (Completed)