Series of single chip SoCs equipped with ARM, FPGA and GPU are available. Linux on ARM/SoC can control FPGA/GPU/Custom-LSI designed for IoT acceleration. In addition, the design-to-test environment for analog-digital-mixed VLSI system is available by integrating the FPGA-SoC and dual-chip ASIC driver.
Series of large-scale FPGA systems on Intel/ARM servers are available. Linux/FreeBSD on Intel/ARM can send large data to FPGA through DMA/PCI interface so that large-scale data to be processed in high-speed.
MATLAB+Simulink and VLISI/FPGA-CAD are available. Circuit/architecture simulation and parallelized applications are performed in high-speed. GPUs such as XTG1080TiOC and V100 are installed.
A stream pool made by JAPAN AQUATECH has been deployed from 2019. The stream pool is approximately 4W x 1.5H x 2D (m). Equipping a swimmer with specific reflective markers, whole-body movement of the swimmer can be tracked by VENUS3D.
This system records the minute ventilation, oxygen intake, carbon dioxide output, heart rate, energy expenditure, and so forth with high accuracy. This lightweight system does not disturb the motion of a subject when he or she wears it for the measurement. We can evaluate multiple physiological aspects under exercise by combining this system with other physiological measurement systems.
With 150 Kilbots and 15 Khepera IVs, we are developing distributed algorithms to cooperatively operate a swarm of robots. Kilbot is a very small (3.3cm) robot that can communicate with nearby Kilobots and move autonomously. Khepera IV is also a small (14cm) robot that can communicate and move. In addition, Khepera IV has many types of sensors (a color camera, infrared sensors, ultrasonic sensors, a microphone, an accelerometer, a gyroscope).
The Baxter is a humanoid, anthropomorphic robot with two 7 DoF arms and state-of-the-art sensing technologies. With force, position, and torque sensors and controllers at every joint, the Baxter robot is suitable for tasks like intelligent assembling and safe collaborative manipulation with humans. Combining with multiple cameras, the Baxter robot can intelligently work in a 3D space while keeping a safe distance to humans to avoid injury.
With versatile arms and precise movements, the Nextage robot (15 DoF, 6 for arms × 2, 2 for neck, 1 for waist) is able to achieve and even surpass the manufacturing ability of most humans. Thanks to its scalable open software environment (based on C++/Python/ROS) and functional body, the Nextage robot is a suitable investigation platform for learning control strategy using modern machine learning/reinforcement learning algorithm.
The UR5 robot has six-axis robot arm with a 5kg payload. Its repeatability is +/- .004″, along with speeds up to 1 meter/sec allowing quick precision handling of even microscopically small parts. It has a great balance between control accuracy and velocity and therefore perfectly matches not only pick up/assembling tasks but also safe collaboration with humans. The UR5 robot supports both simple interactions using intuitive tablet interface and programming on scalable open software environment.
This vehicle provides ad-hoc Internet connectivity using satellite communication facility and its mobility. Therefore, it can be used as emergency Internet environments in case of disasters.
This system can support various application areas such as deep learning, big data analysis, and large scale simulations with computational facilities as follows:
- 54 cluster nodes:
Each nodes is equipped with 24 cores CPU, 256 GiB memory and 1,792 CUDA cores GPU.
- 8 super-parallel calculation nodes:
Each nodes is equipped with 24 cores CPU, 256 GiB memory and 7,168 CUDA cores GPU.
- 2 large shared memory nodes:
Each nodes is equipped with 144 cores CPU and 2 TiB memory.
- 8 large-scale data processing nodes:
Each nodes is equipped with 44 cores CPU, 768 GiB or 256 GiB memory and 96 TB storage for HDFS.
- 1 materials analisys node:
The node is equipped with 44 cores CPU and 768 GiB memory.
- high speed distributed file server:
It works as a GPFS server with 571 TiB capacity and 8 GB/s throughput storage.
This system supports various types of user jobs: batch and interactive jobs, Hadoop, Docker and LXC containers, virtual machine environments. It also provides users with various softwares including several compilers, computation libraries, numerical analysis applications, materials analysis applications.
An open-magnet low-field (0.25 T) MRI (magnetic resonance imaging) system (G-scan Brio, Esaote, Genova, Italy) is available at NAIST. Its main feature is that weight-bearing standing position imaging is possible in addition to a supine position, which is particularly suitable for musculoskeletal imaging under gravity. Computational models of functional anatomy and biomechanics of human musculoskeletal systems are constructed from acquired MR images, which will be applied to medical diagnosis, therapy planning, rehabilitation, and sports science.
This system consists of 27 nodes, 90 GPUs of NVIDIA GeForce TITAN, NVIDIA GeForce GTX1080Ti for deep learning.
This system consists of 7 nodes, 1.6TB Memory, 100TB Disk array, 4 GPUs of NVIDIA Tesla for deep learning.
High-speed and large-memory computing servers
A 7-node computing cluster with a total of 504 cores and 900GB of memory, as well as a high-speed distributed file system capable of holding 110 TB of data, allowing for the real-time acquisition and processing of large-scale data regarding all aspects of communication. The cluster is being used for research into the high-speed and flexible access of text and speech data, as well as multi-lingual text analysis and translation.
This wearable eye tracking system captures gaze behavior in any real world environment. Its wide angle scene camera offers a view from viewpoint of a subject. This system allows researchers to see exactly what a person is looking at. It is possible to synchronize this system with a broad spectrum of physiological data, including EEG, ECG, and motion capture. Since there are two tracking systems, simultaneous recoding from two subjects is also possible.
This screen-based eye tracker captures gaze data at 300 Hz with extremely-high accuracy and precision allowing for head movements. This system includes the subsystem to reliably and accurately synchronize on-screen stimuli and eye tracking data with various physiological data including EEG.
This system measures multi-channel, high resolution electroencephalography (EEG) and surface electromyography (sEMG). A headcap with pin-type active electrodes realizes a quick start of EEG measurement. sEMG signals are measured with a flexible high density array of electrodes pressed on the skin. The array setup provides information about the sEMG topography.
This simple compact driving simulator is used for the research of driving behaviors. This system is used with the glasses-type eye tracking system to investigate the driver’s visual information processing during driving.
This system is used for the research of human movements. Joint torques and muscle tensions in a whole body can be estimated from measured data using a musculoskeletal simulator.
This robot interacts with humans by using not only verbal channel but also non-verbal communication channel such as gesture. This robot is also used for concept learning by using its visual, auditory, and tactual sensors.
These devices enable us to execute spectroscopic analysis widely used in many fields. Hyper spectral camera captures a cubic image of 696pix by 520pix by 128bands across the 400 to 1,000 nm spectral range. Three kinds of spectrometers are also available that cover 200-1,100nm, 500-1,375 nm, and 900-2,500 nm
This facility provides, as a testbed, an actual home environment with a living room, a kitchen, a toilet, a bathroom, a bedroom, where various home appliances are deployed as in an ordinary household. In addition, this facility is equipped with special sensors including high-accuracy indoor positioning system, wireless power meters, door sensors, ambient sensor (temperature, humidity, illuminance, etc), and motion sensors. We are collecting data while subjects are actually living in this facility and are developing various methods including activity recognition and automatic appliance control using the collected sensor data.
A device to measure brain waves and other biometric signals. Using this 32-channel battery-powered brain wave measurement and analysis device, we are performing measurements of the brain activity caused by human communication.
An ultra large screen display system composed of six large hi-vision monitors. This system is available flexibly as a single large display or separated two displays composed of four monitors and two monitors respectively.
A private cloud system integrated by a virtual computing platform for software analysis and cloud computing researches. This system includes an ultra high-speed and reliable network storage providing 90TB of RAID6 HDD and a set of blade servers with 160 CPU cores.
A system composed of different panels with 5 SXGA Plus projectors, 3 high definition plasma display (50 inches), 2 vertical touch panels (42 inches), 1 moving projector system, and a UMU SCREEN. We use this system to study how to improve meetings and human communications.
A high-performance computing system for research on large-scale
natural language processing, including parsing volume of
streaming text in real time, translating multi-lingual documents,
and text mining.
The system consists of a 24-core Xeon E5-4617 large-scale computing server with 512 GB RAM, four 8-core Xeon E5-2643 middle-scale computing servers with 256 GB RAM each, and two network file servers with a total of 44 TB storage.
Robots operating in human-populated environments require the abilities to behave in a human-like manner. Humanoid robots are the most suitable platform to achieve this because of their human-like appearance. This humanoid robot has 34 DOF in its whole body, including 7 DOF in each arm and 6 DOF in each leg. These many DOF enable the robot to execute tasks usually performed by humans in shared spaces. In spite of its high number of DOF, the robot is compact and slim (151 cm in height and 38 kg in weight) and, therefore, superior in both operability and applicability. Our research focuses on making this humanoid robot achieve intelligent whole-body movement.
An imaging system with four times the resolution of high vision images. It is capable of sending / receiving 4K images and displaying realistic images. This system has been used in research on new generation multimedia systems, combining image processing and network technologies, to transmit bulk data regardless of the distance from the remote end through the internal and external information superhighway.
Coexistence of autonomous robots and humans in the same environment presents the challenge of how to achieve new ways of interaction between them. This system enables many sensors to be used to capture human and robot behavior, and real-time recognition can create or initiate interaction. This system consists of a 158-cm high robot with 31 degrees of freedom, multiple cameras and a motion capture system.
日常生活での作業・物体操作を研究するための双腕ロボットシステムです。頭部に装着された 2眼カメラシステムをパン・チルトさせて環境の情報を取得することができ，腕部に６自由度，手先に４自由度を有 する手腕により器用な動作を実現することができます。またOpenRTMと呼ばれるロボット向けミドルウエアを使用することで、システム構成も柔軟に変更することができます。今後、 このロボットシステムを用いて、人の作業知能の解明に取り組んでいきます。
Satellite or wireless communication can connect remote graphics workstations. This is done for research in communication systems and image processing, enabling remote medical diagnostics using VR and videoconference systems. The system comprises an MPEG-2 video image encoder / decoder device, an image processing workstation to generate the VR image, and high-speed communication and mobile communication systems.
A 7-DOF manipulator controlled by 16 pneumatic artificial muscles. Due to its softness, it can be used in assisting people through direct mechanical contact.