Yusuke Yoshiyasu

About

Yusuke Yoshiyasu is a robotics researcher whose work spans robot perception, autonomous navigation, and intelligent motion planning. His research centers on three interconnected domains: sensor fusion and odometry estimation, reinforcement learning-based locomotion and manipulation, and object pose estimation and scene understanding. Yoshiyasu has made notable contributions to learning-based odometry, developing TransFusionOdom, a transformer-based LiDAR-inertial fusion system that has garnered 19 citations, and CertainOdom, which incorporates uncertainty-weighted multi-task learning for safer deployment. His work on quadrupedal and manipulator robotics leverages Riemannian Motion Policies combined with multi-agent reinforcement learning frameworks, enabling robots to perform advanced locomotion and reactive motion generation in dynamic, cluttered environments. He has also explored object goal navigation enhanced by large language models, bridging classical robotics with modern foundation models. Earlier contributions include 6-DOF object pose estimation using CNNs with minimal training data, and the NeuralLabeling toolset, which harnesses Neural Radiance Fields to streamline vision dataset annotation. His cumulative body of work reflects a consistent drive to make robots more capable, adaptive, and safe across perception and control tasks — positioning him as an emerging contributor to the robotics and embodied AI communities.

Research Focus

Key Achievements

Top Papers

1. 1
   TransFusionOdom: Transformer-Based LiDAR-Inertial Fusion Odometry Estimation

   19 citations · 2023
2. 2
   Enhancing multimodal-input object goal navigation by leveraging large language models for inferring room–object relationship knowledge

   8 citations · 2025
3. 3
   Learning Advanced Locomotion for Quadrupedal Robots: A Distributed Multi-Agent Reinforcement Learning Framework with Riemannian Motion Policies

   7 citations · 2024
4. 4
   Deep Reactive Planning in Dynamic Environments

   5 citations · 2020
5. 5
   A Hierarchical Robot Learning Framework for Manipulator Reactive Motion Generation via Multi-Agent Reinforcement Learning and Riemannian Motion Policies

   4 citations · 2023
6. 6
   Toward 6 DOF Object Pose Estimation with Minimum Dataset

   4 citations · 2019
7. 7
   CertainOdom: Uncertainty Weighted Multi-task Learning Model for LiDAR Odometry Estimation

   3 citations · 2022
8. 8
   Understanding and exploiting object interaction landscapes

   3 citations · 2017
9. 9
   NeuralLabeling: A versatile toolset for labeling vision datasets using Neural Radiance Fields

   2 citations · 2024
10. 10
    Scanning and Affordance Segmentation of Glass and Plastic Bottles

    2 citations · 2024

Key Collaborators

Contact & Links