Tokyo University of Science - Department of Mechanical Aerospace Engineering, Arai Laboratory

The Intelligent Robotics Laboratory conducts research and development on robotic systems that address social issues using technologies such as robotics, control theory, machine learning, and image processing. Our work focuses on the key technologies required for robotic manipulation—3D measurement, environment and object recognition, grasp planning, and visual servoing—and their integration into complete robot systems. We are also exploring applications of these technologies in biomedical image analysis and multi-agent systems.

Automation of E-waste Disassembly (Bolt Removal)

This system automatically removes bolts by equipping a robot arm with a camera, projector, and electric screwdriver, which are controlled through vision-based servoing. As the amount of electronic waste continues to increase, improving recycling efficiency has become essential for achieving SDGs. However, since disassembly targets are diverse and low-volume, conventional teaching-and-playback systems using fixed jigs are unsuitable. Therefore, a universal automatic disassembly system that requires no positioning jig is indispensable. Focusing on bolt removal, we achieved automation through active visual servo control of the screwdriver’s positioning.

Automation of E-waste Disassembly (Wire Cutting)

This robot system cuts wires by recognizing the wiring and cutting position using a camera and cutting tool mounted on the robot arm. The recycling rate of electronic waste remains below 50%, as disassembly is still largely manual. To accelerate recycling, we need a general-purpose disassembly system that requires no jigs. Our research focuses on wire cutting—since deformable objects such as cables and harnesses lack CAD data and are difficult to recognize, we achieved automation using instance segmentation and keypoint detection.

Picking of Transparent and Specular Objects

We study 3D reconstruction methods for scenes involving transparent or reflective objects, which are difficult to measure accurately. Using a hand-eye camera, the system captures images from multiple viewpoints, reconstructs the scene in 3D, and performs picking based on the reconstruction results. Because this method relies only on a monocular camera, it can be implemented at low cost. This research was supported by the JKA Foundation.

Bin Picking and Kitting Robot System

This robot system performs bin picking and kitting tasks, which are essential in logistics and manufacturing. It picks up objects piled randomly in a bin and arranges them in a designated location. As e-commerce companies like Amazon and Rakuten pursue warehouse automation, bin-picking robots are becoming indispensable. Similarly, in factories, efficient production under variable demand requires such systems. Our research focuses on key technologies—3D measurement and recognition, grasp planning, motion planning, and active visual servoing—to realize high-performance bin-picking and kitting robots.

Active Visual Servoing

Visual servoing refers to the theoretical framework of controlling a robot based on visual information. Just as humans rely on vision to understand and interact with their environment, autonomous robots require visual feedback to perform tasks accurately. We propose a new framework called Active Visual Servoing, which extends conventional methods and enables high-precision manipulation of textureless or reflective objects.

Object Regrasping

Object regrasping is an essential skill for general-purpose autonomous robots. In this study, we achieved sub-millimeter regrasping accuracy using a dual-arm robot by applying our Active Visual Servoing technique, enabling fast and precise handover between arms.

3D Measurement and Recognition

For autonomous robots, 3D measurement is a fundamental capability, much like human binocular vision. Although various sensors exist, measuring metallic or transparent objects remains difficult. We aim to develop a versatile 3D measurement device capable of accurate and inexpensive sensing for objects with diverse optical properties. Incorporating mathematical approaches such as sparse modeling and deep learning, we pursue both theoretical innovation and practical implementation for industrial use.

Collaborative Robot System

The number of collaborative robots working alongside humans is increasing yearly. To operate efficiently, such robots must predict human movements and act autonomously to shorten task times while ensuring safety and comfort. Our research aims to realize these systems by studying human motion prediction and autonomous motion planning for collaborative robots.

Behavior Analysis of Model Organisms

To effectively combat cancer and infectious diseases, it is essential to understand the control mechanisms underlying various biological phenomena. This study targets the fruit fly, one of the most commonly used model organisms, and proposes image-based identification and tracking algorithms. More than 70% of human disease-related genes exist in Drosophila. Using our method, we clarified the function of taste sensory neurons located in the fly’s legs. Part of this research was published in Nature Communications. By applying this technique, large-scale drug screening using genetically modified Drosophila can be achieved, significantly accelerating the development of new medicines.

Research Achievements

Background

Affiliation: Associate Professor, Department of Mechanical and Aerospace Engineering, Faculty of Science and Technology, Tokyo University of Science

Degree: Ph.D. in Information Science (March 2010, Tohoku University)

Professional Experience:

April 2022 - Present: Associate Professor, Department of Mechanical Engineering, Tokyo University of Science
April 2016 - March 2021: Associate Professor, Graduate School of Engineering, Tohoku University

Awards

December 2023: Best Presentation Award, SI2023
April 2021: Young Scientist Award, Ministry of Education, Culture, Sports, Science and Technology
March 2020: Best Presentation Award, SI2019
June 2019: Best Presentation Award, Robotics and Mechatronics Division
March 2019: Best Presentation Award, SI2018, Society of Instrument and Control Engineers
December 2018: Paper Award (FA Foundation)
October 2018: 32nd Journal Paper Award, Robotics Society of Japan

Main Publications

Optimal Projection Pattern for Active Visual Servoing (AVS)
Shogo Arai et al., IEEE Access 12, 47110-47118 (April 2024)
FPCC: Fast point cloud clustering-based instance segmentation for industrial bin-picking
Yajun Xu, Shogo Arai et al., Neurocomputing 494, 255-268 (2022)
Convolutional Neural Network-Based Visual Servoing for Eye-to-Hand Manipulator
Fuyuki Tokuda, Shogo Arai et al., IEEE Access 9, 91820-91835 (2021)

MISC

Simultaneous Optimization of Manipulator Path and Layout for Automated Assembly
Raito Murakami, Shogo Arai, SI Division Conference, Society of Instrument and Control Engineers, December 2023
Proposed Handling Method for Objects with Diverse Optical Properties using Reinforcement Learning
Ginga Kennis, Takumi Sagano, Shogo Arai, SI Division Conference, Society of Instrument and Control Engineers, December 2023
Active Visual Servo in Hand-Eye System for Automatic Bolt Removal
Kensei Tanaka, Shogo Arai, SI Division Conference, Society of Instrument and Control Engineers, December 2023

Intellectual Property Rights

Patent No. JP2018-189510: Method and Apparatus for Estimating Position and Orientation of 3D Objects
Patent No. JP2018-097889: Object Recognition Device, Object Recognition Method, Object Recognition Program, Robot System, and Robot
Patent No. 6299150: Control Device, Robot, Control System, Control Method, and Control Program
Patent No. JP2018-017653: 3D Measurement Device, Robot, Robot Control Device, and Robot System
Patent No. JP2017-219365: Position and Orientation Calculation Device, Robot Control Device, and Robot

Announcements

Research Topics in Arai Laboratory

Automation of E-waste Disassembly (Bolt Removal)

Automation of E-waste Disassembly (Wire Cutting)

Picking of Transparent and Specular Objects

Bin Picking and Kitting Robot System

Active Visual Servoing

Object Regrasping

3D Measurement and Recognition

Collaborative Robot System

Behavior Analysis of Model Organisms

Members

Shogo Arai

Ryosei Kaneko

Ginga Kennis

Kotoya Hasegawa

Koki Moriya

Shuu Kato

Taisuke Sato

Yurina Tanaka

Keitaro Hiraoka

Riina Matsushita

Shunya Miwa

Yusuke Yamaguchi

Kosei Ito

Momoka Kaneoka

Saki Kumai

Shota Shimazaki

Taiki Hatakeyama

Sakura Hayakawa

Ryota Yagi

Ritsuya Yamamoto

Research Achievements

Background

Awards

Main Publications

MISC

Intellectual Property Rights

Access

Google Map

Campus Map

近日開催予定セミナー

生産性向上ロボット 導入・活用セミナー

生産性向上ロボット導入・活用セミナー