Robot Technologies and Automation Solutions in Digital Era

Robot Technologies Forming Cyber-Physical Systems with Matryoshka-Like Nested Structure
HORI Osamu

Efforts Aimed at Sophistication and Practical Realization of Robot Technologies Needed for Cyber-Physical Systems
MIYAUCHI Takashi / KONDO Koichi / YAMAMOTO Takehiko / YAMAMOTO Hiroshi
Accompanying the growing labor shortage due to the change in the population structure as a consequence of the declining birthrate and aging of society as well as the further diversification and complexity of work, there is a strong need for robots and automated equipment that can perform autonomous operations responding flexibly to the surrounding environment and work requirements in addition to the conventional forms of repetitious work.
The Toshiba Group has been developing many different varieties of robots and automated equipment to meet the needs of society and the market. We are also promoting not only the research and development of more intelligent and advanced technologies for individual robots but also the creation of new value, which has not been able to be achieved independently up to now, by connecting them to cyber-physical systems (CPS) as devices for edge computing aimed at providing automation solutions to a wide range of customers.

Image Recognition Techniques Enabling Precise Understanding for Robots
MATSUMURA Masafumi / ITO Satoshi / AKAGI Takuma
With the rapid progress of artificial intelligence (AI) and deep learning, the advancement and dissemination of robots have been accelerating. The range of robot applications has also been expanding, from the handling of repetitious work at manufacturing sites to various industrial spheres including production, logistics, and physical distribution systems in order to address social issues that have arisen in Japan due to the labor shortage accompanying the aging of society and declining birthrate in recent years. Techniques corresponding to human vision that make it possible to precisely understand operating conditions play a critical role in the sophistication of robots.
The Toshiba Group is engaged in the development of robots equipped with a variety of functions including piece picking, loading and unloading, autonomous mobile, and spot-welding validation robots. In order to enhance the sophistication of these robots, we have developed the following image recognition techniques: (1) an instance segmentation technique that is independent of the type of object, (2) a three-dimensional (3D) posture estimation technique that responds to various shapes of objects, (3) a self-localization technique that achieves an identification accuracy of within 1° in rotation error and within 10 cm in translation error, and (4) an ultrasonic sensing technique that makes it possible to evaluate spot welds nondestructively.

Robot Control Techniques Providing Quick, Careful, and Secure Object Gripping Capability
HIRAGURI Kazuma / KAWAI Hirofumi / NAKAMOTO Hideichi
The range of application of industrial robots has been expanding in recent years from simple tasks at assembly lines to non-routine tasks involving the handling of various objects in a diverse array of fields including the logistics and service areas. Moreover, automation systems that can perform various actions to meet user needs through a combination of robots produced by different companies as required are now attracting attention.
With these trends as a background, Toshiba Corporation has developed a controller interface (I/F) software technique that makes it possible for robots having different types of setup procedures and command systems to operate in a unified manner by means of a generalized communication I/F. We have also developed the following mechanisms: (1) compound mechanisms by which an object can be either attached to or gripped as required, and (2) a robotic hand incorporating mechanisms that allow it to either attach to a lightweight object or support a heavy object from below according to the weight, thereby achieving quick, careful, and secure gripping of various objects using a robotic arm.

Software Platforms Supporting Efficient Development of Various Types of Robot Systems
HIRAYAMA Noriyuki / SAWA Kazuhide / NOGUCHI Keiko
The introduction of robots in various fields for automation and labor saving continues to progress, making it necessary to supply various types of robot systems to individual customers swiftly and at lower cost in order to meet their diversifying requirements. There is consequently an increasing need for the construction of software platforms to support the efficient development of software for robot systems.
As part of its efforts in the development of robots positioned as key edge computing devices in cyber-physical systems (CPS), the Toshiba Group has defined a reference architecture common to robot systems. By not only making best use of the Robot Operating System (ROS) as an open-source software framework but also effectively utilizing our proprietary functions for robot applications, we have developed optimal software platforms for picking robots and autonomous mobile robots, respectively.

Picking Robots and Palletizing Robots Contributing to Labor Saving and Efficient Operation of Physical Distribution Centers
OGAWA Akihito / USHIYAMA Takafumi / EHARA Koji
In the logistics business field, there is a strong need for investments in the construction of physical distribution centers due to the rapid expansion in sales of large volumes of various types of products worldwide through electric commerce. As most of the work at distribution sites, such as processes for sorting and packing, is still dependent on human resources, demand for the introduction of automated processes using robots has been increasing in order to solve a critical labor shortage.
To meet customers’ requirements for labor-saving or totally automated physical distribution centers, Toshiba Infrastructure Systems & Solutions Corporation has developed a picking robot and a palletizing robot following its previous launching of a depalletizing robot. These robots are cyber-physical systems (CPS) that can autonomously judge the on-site situation and perform appropriate work on various objects by repeating a cycle of recognizing the surrounding conditions based on sensor information, generating an operation plan, and executing the operation. Furthermore, through the connection of these robots as edge computing devices to a total warehouse control system (TWCS), this high-level CPS can be expected to realize efficient supervision of the entire site.

Spot Welding Inspection Robot Achieving Labor Saving and Improvement of Reliability Using 3D Ultrasonic Inspection Equipment
USHIJIMA Akira / SAITO Masahiro / MATSUMOTO Shin
Spot welding is an important process in the automobile and rolling stock fields, closely related to the strength of joints. However, conventional manual destructive inspection methods using sampled products are still the mainstream in the process of manufacturing these products, which typically have several thousand welded parts. Accompanying the widespread use of high-tensile-strength steel plates in these fields, demand has now arisen for nondestructive methods allowing such welded parts to be automatically and swiftly inspected with high accuracy.
The Toshiba Group has developed a spot welding inspection robot integrating its Matrixeye three-dimensional (3D) ultrasonic inspection equipment based on a synthetic aperture focusing technique and its propriety robot control techniques. This robot makes it possible to adjust the tilt angle of the ultrasonic probe of Matrixeye, which requires highly skilled operation of the end effector of the robot, in a short time by applying ultrasonic image analysis and precise robot positioning control techniques. We are aiming to achieve the practical application of this spot welding inspection robot at an early stage by improving its reliability and usability through repeated field tests.

Autonomous Mobile Robot System Allowing Swift Development
YAMAMOTO Daisuke / KISHI Nobuyuki / TAIRA Kazuki
Robots are a focus of expectations as a replacement for conventional manual operations in order to solve the severe labor shortage in a broad range of fields in Japan. However, the need for sophisticated expertise and reduction of the time required to develop appropriate robots for various tasks are significant issues.
The Toshiba Group has been making efforts to develop system technologies for autonomous mobile robots, composed of hardware units for respective functions and a software platform, targeting the following two tasks: (1) the task of transporting carts at a physical distribution center, and (2) the task of checking shelf space at a retail store. We have confirmed the effectiveness of our autonomous mobile robot system through demonstration experiments applying prototype robots developed in a short period of time using this system.

Automated Guided Vehicles Equipped with Robot Arm Realizing Labor Saving through Collaboration between Workers and Robots
TERADA Akihito / HORIE Kazuhiro / HARA Naoyuki
The automation of manufacturing sites is becoming an urgent issue in Japanese manufacturing industries as a result of the domestic labor shortage as well as wage increases centering around Asian countries. In particular, automated conveyance systems that can provide not only transport but also parts handling functions are expected to contribute to a reduced need for factory workers when investing in facilities.
In response to this situation, the Toshiba Group has been working on the development of automatic guided vehicles (AGVs) equipped with different robot arms to perform various types of functions and has been introducing these AGVs to manufacturing sites both within and outside the company. We have now developed an AGV equipped with a robot arm to transport semiconductor wafer cases for semiconductor lines, and a mobile palletizer to perform the task of stacking cardboard boxes on pallets for production lines of general sites. These AGVs contribute to the enhancement of worker safety on the production line by monitoring workers using sensors so as to secure a safe distance from them while operating.

CPS Edge Tool Capable of Connecting Existing Equipment to IoT Networks
KUSAKA Tsubasa / TAKI Toshikazu / LIN Chiayu
The collection of a wide variety of data from robots and legacy automated equipment already in operation at manufacturing sites via networks is essential when converting existing facilities into an optimal cyber-physical system (CPS). Demand has therefore been growing for a mechanism that can be attached to equipment lacking a network function in order to easily connect such equipment to Internet of Things (IoT) networks.
Toshiba Corporation has developed a CPS edge tool as an add-on tool that can easily collect data from equipment already in operation. The CPS edge tool incorporates a method for collecting various types of sensor data by assigning ID numbers to each of the sensors, and a meter recognition method to extract numerical values from images of the displays of non-digitalized meters obtained by cameras. This tool is expected to achieve labor saving and make it possible to introduce inspection systems using mobile robots equipped with cameras.

DNN Compaction Method Eliminating Zero Weight Coefficients
YAGUCHI Atsushi / ASANO Wataru / TANIZAWA Akiyuki
Multilayer neural networks for deep learning, or so-called deep neural networks (DNNs), deliver superior performance in various applications including image recognition. However, as DNNs often tend to increase in scale and complexity, requiring higher processing performance, it is difficult to implement applications using them on edge devices with limited computation power and memory capacity.
While investigating how to make such large-scale DNNs more compact, the Toshiba Group discovered a phenomenon in which a portion of the large number of weight coefficients of a DNN automatically converge to zero when training a DNN under general conditions. In cooperation with the Institute of Physical and Chemical Research (RIKEN), we have elucidated the principle of this phenomenon and developed a DNN compaction method to reduce the volume of data in a DNN by eliminating those zero weight coefficients after training. Experiments on image recognition using open datasets have verified that this method achieves a reduction in the weight coefficients of a DNN of more than 80% compared with conventional methods while maintaining recognition accuracy.

DC Power Supply Equipment for Substations Using SCiB™ Lithium-Ion Rechargeable Batteries
TOYOSAKI Tomohiro / KAWAMOTO Shinya / KIUCHI Masako
Although lead-acid batteries are still the mainstream of batteries for DC backup power supply equipment at the substations of electric power companies, the need for a large installation area as well as maintenance costs for periodic inspections and early replacement compared with other equipment are critical issues. Demand has therefore been growing for the introduction of lithium-ion rechargeable batteries that can achieve both space saving and high maintainability. However, it is necessary to ensure the safety of backup power supply equipment used in this field and prevent deterioration of the capacity retention rate during float charging.
The Toshiba Group has developed new DC power supply equipment using its SCiB™ lithium-ion rechargeable batteries offering a high level of safety and a long lifetime. Experiments on SCiB™ modules and full-scale demonstration tests using prototype equipment have verified that the capacity retention rate shows negligible deterioration during float charging. We are promoting the introduction of SCiB™ DC power supply equipment in electricity markets.

Application of Metal Additive Manufacturing Method to Steam and Gas Turbine Materials
HINO Takehisa / OHNISHI Haruki / TADANO Satoshi
Metal additive manufacturing (AM) is attracting attention as a technology that makes it possible to produce metal parts in a short period of time without die making through a process of repeated melting and solidification of metal powders using an electron or laser beam as a heat source. This technology has begun to be applied to various fields. The need has also arisen in recent years for reductions in the number of parts in industrial equipment, such as components for power generation plants, through the use of integrally molded parts, and for increased design flexibility through the ability to manufacture parts having complicated shapes.
Toshiba Energy Systems & Solutions Corporation is engaged in the development of metal AM technologies for application to equipment for thermal power plants, including steam and gas turbines. We have selected a powder bed fusion method for this purpose due to its high shaping speed and accuracy. As a result of the optimization of raw metal powders and process parameters for high-temperature materials to be used in turbine blades, we have confirmed the fine metallurgical structure and good mechanical characteristics of such materials through mockups. We have also developed a simulation method that allows optimal support structures to be designed in order to suppress deformation in metal AM processes.

Flowing-Water Sterilization Equipment Using Environmentally Friendly Deep-UV LEDs
SAKURAI Naoto / KATO Takeo / TAUCHI Akihiko
The mainstream methods for water sterilization still require either treatment with chlorine or irradiation with ultraviolet (UV) light emitted by mercury lamps. The environmental effects of chlorine itself and of mercury spilled from broken lamps have become a critical issue in recent years. In this context, attention is being increasingly focused on a deep-UV light-emitting diode (LED) as a mercury-free UV light source with high safety and increased design flexibility due to its small size despite its low luminous efficiency, not only as a replacement for current mercury lamps but also for various applications in new markets.
Toshiba Lighting & Technology Corporation has developed flowing-water sterilization equipment using 280 nm deep-UV LEDs to satisfy the requirements of the food and medical fields for high levels of safety and security. Through the application of our proprietary flow path design method that makes it possible to effectively irradiate flowing water with deep-UV light, this equipment has achieved an Escherichia coli sterilization ratio of more than 99.9% by means of one-path sterilization treatment at a water flow rate of 10 L/min.