A5088575863- Robot Manipulation and Learning
- Robotic Path Planning Algorithms
- Reinforcement Learning in Robotics
- Human Motion and Animation
- Hydraulic and Pneumatic Systems
- Robotic Locomotion and Control
- Autonomous Vehicle Technology and Safety
- Traffic Prediction and Management Techniques
- Catalytic C–H Functionalization Methods
- Image Processing and 3D Reconstruction
- Artificial Intelligence in Games
- Robotic Mechanisms and Dynamics
- Educational Robotics and Engineering
- Generative Adversarial Networks and Image Synthesis
- Oxidative Organic Chemistry Reactions
- Radical Photochemical Reactions
- Transport Systems and Technology
- Advanced Sensor and Control Systems
Waseda University
2020-2023
University of Fukui
2013
Kayaba Industry (Japan)
2005
Photoinduced electron transfer (PET) promoted decarboxylation of α-(ω-carboxyalkyl) β-keto esters undergoes radical ring expansion and cyclization reactions. This mild environmentally friendly method can provide one-carbon expanded γ-keto bicyclic alcohols, the product distribution is strongly dependent on length alkyl chain containing terminal carboxylate group.
We propose a new method for collision-free planning using Conditional Generative Adversarial Networks (cGANs) to transform between the robot's joint space and latent that captures only areas of space, conditioned by an obstacle map. Generating multiple plausible trajectories is convenient in applications such as manipulation robot arm enabling selection avoids collision with or surrounding environment. In proposed method, various avoid obstacles can be generated connecting start goal state...
In this study, we conduct the first trial ever to realize a robotic buttoning task using dual-arm robot. A robot must handle flexible object (clothes) and solid objects (buttons) simultaneously during task, therefore there is no previous work due its complexity. We design strategy of by dividing series motions into subtasks with following methods: (a) marker-based algorithmic method, markerless machine learning methods (b) without pseudo-rehearsal (c) motions. The consolidation generated...
Human often cross their legs unconsciously while sitting, which can lead to issues like shifting the center of gravity, lower back pain, decreased blood flow and pelvic distortion. Detecting unconscious leg-crossing is important for maintaining correct posture. In this study, we explored detection postures using machine learning on data from body pressure distribution sensors. Collected 180 seconds 4 male subjects (25.8 ± 6.29 y.o.) in three conditions: no leg crossing, right crossed, left...
We show a new method for collision-free path planning by cGANs mapping its latent space to only the areas of robot joint space. Our simply provides this after which any planner, using optimization conditions, can be used generate most suitable paths on fly. successfully verified with simulated two-link arm.
No matter their environment, it is important for robots to ensure safety and prevent malfunctions. An aspect robotics therefore collision-free path planning. Conventional methods such as RRT potential field have difficulty adapting dynamically changing environments, high calculation cost iterating or searching. In this research, we propose a new planning method in which the latent space of Conditional Generative Adversarial Networks (cGANs) represents workspace where robot avoids obstacles,...
We propose a new method for collision-free planning using Conditional Generative Adversarial Networks (cGANs) to transform between the robot's joint space and latent that captures only areas of space, conditioned by an obstacle map. Generating multiple plausible trajectories is convenient in applications such as manipulation robot arm enabling selection avoids collision with or surrounding environment. In proposed method, various avoid obstacles can be generated connecting start goal state...
In robot motion, collision avoidance is essential to ensure safety, and generation of multiple paths according the objective beneficial. Conventional methods have been pointed out difficulty in achieving both optimization, large computational costs. It also challenging compensate for safety using deep learning methods. this paper, we propose a method that can generate an optimal trajectory with obstacles short time guarantee avoidance. The calculated by updating latent variables posture...
In this study, we propose a robotic buttoning task by dual-arm humanoid robot using marker-based recognition. our method, design dividing into multiple subtasks, and the performs them in order. When performed each subtask, it conducts adjusts motions based on position of button its hole obtained maker-based For implementation task, used Baxter, an experimental T-shirt, button-aid. As result experiments, although problem low success rate remains, was able to realize series operations.