I am researching manipulation robots equipped with physical embodiment and intelligence capable of robust and agile operations in the real world, to create the foundation for next-generation industries.
This study presents a gait generation method for legged modular robots with easily detachable joints. The approach balances joint load and mobility, ensuring stable and efficient locomotion. This is particularly useful for optimizing the complex movement constraints and desired behavior of multi-legged robots, such as 4-legged and 6-legged robots, during physical simulations before real-world deployment.
This study proposes a misalignment-tolerant tool changer for modular robots with easily detachable joints. Instead of complex force sensors, it relies on passive self-alignment geometries, such as chamfered receptacles and triangular guides, to correct positioning errors. Paired with a compact rotating exchange station, real-world tests confirm the system effectively absorbs execution errors, enabling autonomous tool reconfiguration.
This research proposes a framework for replanning human-robot collaborative tasks using Vision-Language Models (VLMs). By employing a dual-correction mechanism that addresses both semantic understanding and physical execution errors, the system achieves the generation of tasks aligned with human intent based solely on ambiguous language instructions and visual observations, successfully executing collaborative scenarios such as part fastening and tool handovers.
A soft jig inspired by a jamming gripper enables the flexible regrasping of diverse parts. By forming a cavity within the membrane to temporarily and stably hold the target object, it achieved a placement success rate of over 80% in drop-placement experiments, demonstrating its potential as a practical alternative to rigid fixtures.
This study presents a shell-type soft jig for robotic disassembly that securely holds diverse objects without causing damage. Its balloon-based mechanism wraps around varying shapes, enabling part extraction without dedicated jigs or precise planning. Feasibility was demonstrated through experiments on ten objects, comparing it against a vise and a jamming soft jig.
This study introduces a hierarchical optimization method for reconfigurable robotic disassembly of constrained structures. Integrating a linear-rotary stage and multi-arm robots with diverse tools, it combines multi-objective genetic algorithms for sequence-task-motion planning and constraint programming for scheduling to realize integrated planning for long-horizon operations.
This study presents an affordance-guided teleoperation system for dual-arm disassembly of mating parts. By visualizing feasible grasps and disassembly directions in a virtual environment, it improves operability despite occlusions. A hybrid position-impedance controller prevents excessive tracking under load, achieving higher success rates and reduced pose deviation in real-world experiments.
A modular vacuum-based fixturing system using balloon-type soft grippers conforms to complex, curved surfaces of small appliances, providing stable support for screw removal. A stability-aware gripper placement planning framework samples and evaluates contact points via convex hull-based criteria. Experiments show consistently higher success rates and placement stability compared to traditional rigid fixtures.
This work presents an LLM-based Task and Motion Planning framework that generates executable cooking task plans from videos with subtitles. By integrating Functional Object-Oriented Networks (FOON) to validate and refine plans, the system mitigates LLM hallucination and video uncertainty. In tests on five recipes with a dual-arm robot, four plans from our method were successfully executed, compared to only one using an LLM-only approach.
This study enables mobile grasping for commercial robots using self-supervised learning to adjust velocity and grasp based on object shape. It simplifies the task into three action primitives, reducing data sparsity. Three FCNs predict grasp actions and correct motion errors. A two-stage learning approach improves accuracy, and randomized simulations enhance generalization across various objects and environments.
This study proposes a pick-and-toss (PT) method as an efficient alternative to pick-and-place (PP), extending a robot's range. While PT enhances object arrangement efficiency, placement conditions affect toss accuracy. To optimize this, we suggest selecting PP or PT based on task difficulty derived from the environment. Our method combines self-supervised learning for toss motion and a brute-force search for task determination. Simulations and real-world tests on arranging rectangular objects validate this approach.
We tackle many-objective optimization for planning uncertainty-aware sequences and motions in assembling complex mechanical products with many contact points. It takes CAD models as input, simulates assembly, and employs a many-objective genetic algorithm to optimize the order, placement, transitions, grasps, and trajectories for robot execution simultaneously.
The perceptive soft jig extended in this study is equipped with a hydraulic drive system and enables parts-fixing by creating a jammed state while maintaining optical transparency, thereby facilitating visual sensing of the jig's membrane from camera sensors embedded in the jig. Furthermore, we proposed a sensing method to estimate the fixed object pose based on the behaviors of markers created on the jig's inner surface.
We introduce a similarity matching method between novel objects and known objects in a database based on category-association to achieve pick-and-place tasks with high accuracy and stability. We calculate category name similarity using word embedding to quantify the semantic similarity between the categories of the known and target real-world objects. Using a similar model identified by a similarity prediction function, we preplan a series of robust grasps and imitate them to plan new grasps for real-world target objects.
To achieve robust detection and agile manipulation for robotic waste sorting, we propose three methods: graspless manipulation, automated image dataset collection, and appearance gap mitigation. To prevent performance degradation in trained detectors, we address illumination and background discrepancies between training and sorting scenes using computer vision techniques.
Aiming to generate easy-to-handle assembly sequences for robotic assembly, we propose a multiobjective genetic algorithm to balance several objectives for generating constraint-satisfied and preferable assembly sequences. Furthermore, we developed a method of extracting part relation matrices using 3D computer-aided design (CAD) models.
To design a flexible assembly system that can handle objects of various shapes, we propose a jamming-gripper-inspired soft jig that is capable of deforming according to the shape of assembly parts. The soft jig has a flexible membrane made of silicone, which has a high friction, elongation, and contraction rate to keep parts firmly fixed. The inside of the membrane is filled with glass beads to achieve jamming transition.
Training deep-learning-based vision systems requires time-consuming and laborious manual annotations. To automate the annotation, we associate one visual marker with one object and capture them in the same image. However, if an image showing the marker is used for training, the neural network normally learns the marker as a feature of the object. By hiding the marker with a noise mask, we succeeded in reducing erroneous learning.
Assistant Professor (2023-present)
Agile Reconfigurable Manipulation Robots
Visiting Researcher (2023-2024) -
Institute of Robotics and Mechatronics, German Aerospace Center (DLR)
CAD-Informed Uncertainty-Aware Robotic Assembly
Specially-Appointed Assistant Professor (2021-2023)
Novel Object Manipulation
Cross-Appointment Specially-Appointed Assistant Professor (2022-2023) - Robot Learning Laboratory in NAIST
Soft Robotic Assembly
Cross-Appointment Specially-Appointed Assistant Professor (2021-2022) - Human Robotics Laboratory in NAIST
Quickly Deployable Waste Sorter
Ph.D. Student (2018-2021)
Agile Reconfigurable Robotic Assembly System
Research Intern (2018-2020) -
Microsoft Development Applied Robotics Research Team
Interactive-Learning-from-Observation (contributed to LabanotationSuite)
Super Creator (certified by METI & IPA) -
MITOU program 2018
Quickly Deployable Image Recognition AI
Master Course Student (2016-2018)
Tactile-Based Pouring Motion Inspired by Human Skill
Advanced Course Student (2014-2016)
Modeling Forearm Pro-supination for Baseball Pitching Analysis
Technical College Student (2009-2014)
Artificial-Muscle-Driven Robotic Arm with Biarticular Muscles