TKTakuya Kiyokawa

Takuya Kiyokawa

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Takuya Kiyokawa has a total of 53 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Component Influence-Driven Fastener Reduction for Robotic Disassemblability-Aware Design Simplification

    by Takuya Kiyokawa, Tomoki Ishikura, Shingo Hamada, Genichiro Matsuda, Kensuke Harada

    Robotic disassembly must be considered during the product design phase to accelerate automated remanufacturing. However, designers currently lack quantitative feedback for identifying the structural elements that hinder robotic operations. To address this, this study proposes an analytical framework that provides actionable redesign guidance focused on fastener reduction, because fasteners are numerous and ubiquitous components found in almost all manufactured products. Using a computer-aided design (CAD) model and its automatically generated contact-connection-constraint (CCC) graph, the framework translates the robotic disassembly sequence planning outcomes into component influence scores. These scores reflect the frequency with which a component causes structural constraint violations or evaluation objective deteriorations in a robotic disassembly sequence. To visually highlight structural hindrances, the framework projects these scores onto the CAD geometry as 3D heatmaps. The system then analytically simulates the removal of highly influential fasteners. It reports the expected reductions in structural constraints, tool changes, and robot travel distances while preventing structurally unsafe modifications by evaluating geometric stability metrics. Experiments on seven household appliances demonstrated that the framework successfully targeted redundant fasteners. Removing the recommended fasteners simplified the structural dependencies by eliminating 8–132 structural constraints on the graph depending on the structural configuration of each product. Furthermore, it improved the robotic operational efficiency by eliminating unnecessary tool change operations and shortening travel distances by 165–1675 mm wherever it was structurally permissible.

  • Prompt2Craft: Generating Functional Craft Assemblies With LLMs

    by Vitor Hideyo Isume, Takuya Kiyokawa, Natsuki Yamanobe, Yukiyasu Domae, Weiwei Wan, Kensuke Harada

  • Planning 6-D Grasps for Two-Finger Grippers Considering Edge and Corner Contacts

    by Zhengtao Hu, Ruishuang Liu, Mocong Qiu, Hao Chen, Liang Du, Takuya Kiyokawa, Weiwei Wan, Jianjun Yuan, Kensuke Harada

    This article presents a novel model-based algorithm for planning antipodal grasps, especially considering the edge and corner contacts. Conventional model-based planners search for face contacts to build grasp synergies, which ensure superior grasp stability, but inherently restrict the geometric versatility of feasible grasp configurations. With soft contact, edges and corners can reshape fingerpads to form equivalent face contacts, enhancing the ability to resist gravity. This allows candidates with greater configuration variability, rather than being constrained solely by face normals. The developed planner detects regions of faces, edges, and corners from a computer-aided design (CAD) model, samples contacts from each region, detects paired contacts using contact detectors corresponding to contact types, and screens the pairs by stability and collision checking. It generates grasp candidates with qualified stability and flexible configurations, breaking the limits of planning for polyhedra, asymmetric objects, and objects with various sharp angles, etc. Experimental results quantitatively demonstrate that the proposed method has superior performance on grasp distributions and success rates using the selected objects from open datasets and off-the-shelf items.

  • Open Access

    Many-objective-optimized semi-automated robotic disassembly sequences

    by Takuya Kiyokawa, Kensuke Harada, Weiwei Wan, Tomoki Ishikura, Naoya Miyaji, Genichiro Matsuda


  • Open Access

    Robotic Paper Wrapping by Learning Force Control

    by Hiroki Hanai, Takuya Kiyokawa, Weiwei Wan, Kensuke Harada

    Robotic packaging using wrapping paper poses significant challenges due to the material’s complex deformation properties. The packaging process itself involves multiple steps, primarily categorized as folding the paper or creating creases. Small deviations in the robot’s arm trajectory or force vector can lead to tearing or wrinkling of the paper, exacerbated by the variability in material properties. This study introduces a novel framework that combines imitation learning and reinforcement learning to enable a robot to perform each step of the packaging process efficiently. The framework allows the robot to follow approximate trajectories of the tool-center point (TCP) based on human demonstrations while optimizing force control parameters to prevent tearing or wrinkling, even with variable wrapping paper materials. For training, we used a dataset of 480k trajectories expanded from 10 human demonstration trajectories for each subtask. Additionally, position control parameters are trained in the simulation environment for 110k steps, followed by the training of force control parameters on the real robot for 100k steps. The proposed method was validated through ablation studies, which demonstrated that the task success rate improved significantly from 12.5% to 95%, while the rates of tearing and wrinkling were reduced from 51% to 5% and from 52% to 5%, respectively. Furthermore, the force control strategy proved to be adaptable across different wrapping paper materials and robust against variations in the size of the target object.


  • A Multilevel Similarity Approach for Single-View Object Grasping: Matching, Planning, and Fine-Tuning

    by Hao Chen, Takuya Kiyokawa, Zhengtao Hu, Weiwei Wan, Kensuke Harada

    Grasping unknown objects from a single view has remained a challenging topic in robotics due to the uncertainty of partial observation. Recent advances in large-scale models have led to benchmark solutions such as GraspNet-1Billion. However, such learning-based approaches still face a critical limitation in performance robustness for their sensitivity to sensing noise and environmental changes. To address this bottleneck in achieving highly generalized grasping, we abandon the traditional learning framework and introduce a new perspective: similarity matching, where similar known objects are utilized to guide the grasping of unknown target objects. We newly propose a method that robustly achieves unknown-object grasping from a single viewpoint through three key steps: 1) Leverage the visual features of the observed object to perform similarity matching with an existing database containing various object models, identifying potential candidates with high similarity; 2) Use the candidate models with pre-existing grasping knowledge to plan imitative grasps for the unknown target object; 3) Optimize the grasp quality through a local fine-tuning process. To address the uncertainty caused by partial and noisy observation, we propose a multi-level similarity matching framework that integrates semantic, geometric, and dimensional features for comprehensive evaluation. Especially, we introduce a novel point cloud geometric descriptor, the C-FPFH descriptor, which facilitates accurate similarity assessment between partial point clouds of observed objects and complete point clouds of database models. In addition, we incorporate the use of large language models, introduce the semi-oriented bounding box, and develop a novel point cloud registration approach based on plane detection to enhance matching accuracy under single-view conditions. Real-world experiments demonstrate that our proposed method significantly outperforms existing benchmarks in grasping a wide variety of unknown objects in both isolated and cluttered scenarios, showcasing exceptional robustness across varying object types and operating environments.


  • Open Access

    Efficiently Collecting Training Dataset for 2D Object Detection by Online Visual Feedback

    by Takuya Kiyokawa, Naoki Shirakura, Hiroki Katayama, Keita Tomochika, Jun Takamatsu

    Training deep-learning-based vision systems requires the manual annotation of a significant number of images. Such manual annotation is highly time-consuming and labor-intensive. Although previous studies attempted to eliminate the effort required for annotation, the effort required for image collection was retained. To address this issue, we propose a human-in-the-loop dataset-collection method using a web application. To counterbalance workload and performance by encouraging the collection of multi-view object image datasets enjoyably, thereby amplifying motivation, we propose three types of online visual feedback features to track the progress of the collection status. Our experiments thoroughly investigated the influence of each feature on the collection performance and quality of operation. These results indicate the feasibility of annotation and object detection.


  • Open Access

    Synergy Hand Using Fluid Network: Realization of Various Grasping/Manipulation Styles

    by Kazuki Higashi, Keisuke Koyama, Fanny Ficuciello, Ryuta Ozawa, Takuya Kiyokawa, Weiwei Wan, Kensuke Harada

    Synergy hands absorb the complexity of control and mechanisms inherent in multi-fingered hands, enabling human-like grasping with simple mechanisms and low-dimensional control. However, they struggle with functional object grasping and dexterous manipulation expected of multi-fingered hands. In this study, we propose a synergy hand using a fluid network that can balance functional grasping and manipulation with low-dimensional control. The fluidic network comprises actuators attached to the joints of the hand, a single hydraulic pump, connecting flow tubes, and solenoid valves attached at various points in the flow tubes. Switching the network configuration via solenoid valves makes it possible to choose whether each joint is under synergy control or remains fixed while pressurized. This allows for allocating grasping and manipulation functions to each finger, enabling control appropriate to the function. Furthermore, by controlling the flow rate to each actuator with proportional control solenoid valves, the system can precisely express various grasping postures using a single pressure source. Experimental validation of the proposed synergy hand demonstrated its ability to grasp objects of various shapes. The use of proportional control valves for joint position control resulted in about 1.5 % improvement in accuracy compared to the method using standard solenoid valves. Itwas also confirmed that the system could functionally grasp and manipulate objects with degrees of freedom, which had been difficult with conventional synergy hands.


  • Open Access

    Assembly Task Allocation for Human–Robot Collaboration Considering Stability and Assembly Complexity

    by Zhenting Wang, Takuya Kiyokawa, Natsuki Yamanobe, Weiwei Wan, Kensuke Harada

    This paper proposed a task allocation method for human-robot collaborative assembly from a graphical instruction manual based on the assembly complexity while considering the stability of each sub-assembly task. Without using the CAD model of each part to construct the liaison graph, we generate the liaison graph of a product from the 2D graphical instruction manual. To realize a successful assembly task, we especially focus on the support relation of parts considering the stability during the construction of each connection. We check the stability of each sub-assembly task and then decide the timing to add support to maintain stability and the agent to execute this support motion. In addition, to complete an assembly task with additional support motions by a human operator and a robot, other implicit motions need to be considered, such as the direction change of the subassembly and the tool change motion, which is necessary to execute the planned task sequences. The dynamic liaison complexity for each sub-assembly is computed separately for the human operator and the robot to assign different sub-assembly tasks to appropriate operators. The allocation result is sent to a human-centered HRC interface to realize the communication between human operators and robots during the assembly task. Finally, a human operator and a dual-arm robot collaborate to accomplish two chair assembly tasks.


  • Component Selection for Craft Assembly Tasks

    by Vitor H. Isume, Takuya Kiyokawa, Natsuki Yamanobe, Yukiyasu Domae, Weiwei Wan, Kensuke Harada

    Inspired by traditional handmade crafts, where a person improvises assemblies based on the available objects, we formally introduce the Craft Assembly Task. It is a robotic assembly task that involves building an accurate representation of a given target object using the available objects, which do not directly correspond to its parts. In this work, we focus on selecting the subset of available objects for the final craft, when the given input is an RGB image of the target in the wild. We use a mask segmentation neural network to identify visible parts, followed by retrieving labeled template meshes. These meshes undergo pose optimization to determine the most suitable template. Then, we propose to simplify the parts of the transformed template mesh to primitive shapes like cuboids or cylinders. Finally, we design a search algorithm to find correspondences in the scene based on local and global proportions. We develop baselines for comparison that consider all possible combinations, and choose the highest scoring combination for common metrics used in foreground maps and mask accuracy. Our approach achieves comparable results to the baselines for two different scenes, and we show qualitative results for an implementation in a real-world scenario.