Research Topics 2018-19


Signal processing/CMOS integrated circuit technologies for ultra-compact 3D depth-imaging system

Explanation

There is strong demand to reduce cost and size of 3D depth-imaging systems, including LiDAR, for use in the application of ADAS, drone, and precise-motion robotics. We are seeking a Fellow with an innovative idea for realizing a long-range solid-sate LiDAR and a passion for a cause to implement it. The aim of this programme is to demonstrate the proposed idea through the optimized design, implementation and evaluation of the system by using the design-to-production flow of TOSHIBA original photonic-device-mixed CMOS and MEMS processes.

Knowledge and Skills Required

Candidates should possess an expertise in (1) signal processing or (2) integrated circuit design for imaging system (ex. CMOS image sensor, ToF camera, certain depth sensor). More specifically, (1) RTL design, system design and evaluation using FPGA, MATLAB, C, and python (2) CMOS integrated (Mixed-Signal) circuit design, the experience for Cadence Tools.

Related Papers

[1] K. Yoshioka, A. Sai, et al, “A 20ch TDC/ADC hybrid SoC for 240× 96-pixel 10%-reflection< 0.125%-precision 200m-range imaging LiDAR with smart accumulation technique,” ISSCC Digest Technical Papers, p.92-93, Feb. 2018.
[2] A. Sai, et al., “A 5.5mW ADPLL-based receiver with hybrid-loop interference rejection for BLE application in 65nm CMOS,” IEEE Journal of Solid-State CircuitS, Vol.51, No.12, pp. 3125-3136, Dec. 2016
[3] A. Sai, et al., “A 65nm CMOS ADPLL with 360μW 1.6ps-INL SS-ADC-Based Period-Detection-Free TDC,” ISSCC Digest Technical Papers, pp. 336-337, Feb. 2016.
[4] H. Okuni, A. Sai et al., “A 5.5mW ADPLL-based receiver with hybrid-loop interference rejection for BLE application in 65nm CMOS,” ISSCC Digest Technical Papers, pp. 436-437, Feb. 2016.
[5] A. Sai, et al, "A Digitally Stabilized Type-III PLL Using Ring VCO with 1.01psrms Integrated Jitter in 65nm CMOS," ISSCC Digest Technical Papers, pp.248-250, Feb. 2012.
[6] A. Sai, et al, "A Digitally Stabilized Type-III PLL Using Ring VCO with 1.01psrms Integrated Jitter in 65nm CMOS," ISSCC Digest Technical Papers, pp.248-250, Feb. 2012.
[7] A. Sai, et al, "A 570fsrms Integrated-Jitter Ring-VCO-Based 1.21GHz PLL with Hybrid Loop," ISSCC Digest Technical Papers, pp.98-100, Feb. 2011.

APPLY HERE


Meta-learning for designing neural network architectures

Explanation

Deep Learning has shown remarkable results in recent years in a wide range of computer vision tasks and novel network architectures have been proposed, However, with requirements for increased computational power and memory, it has become difficult to operate them using on-board LSI devices or edge devices.  In this research, we are looking for a researcher who can use meta learning approaches to develop an automated model which can operate a deep neural network in an environment with limited computational resource such as an edge device, based on given limits. For example, we are looking for candidates who have brilliant vision and advanced skills to aim to build an automated deep neural network model architecture logic which based on limited expert data can complete a task. Additionally, using our facilities, software and data for research of the automated design of a neural network architecture, candidates will also explore the design of loss functions and optimization methods, and have the opportunity to work on their own related theme.

The aim of this research is to build the basics of a strong technology for Toshiba in the field of logistics, infrastructure, and self-driving cars.

Knowledge and Skills Required

Candidates are required to have a profound knowledge of at least one of the following fields: pattern recognition, machine learning, deep learning, reinforcement learning and robotics. They should have software implementation skills such as C, C++, C#, MATLAB(R) or python(TM).

Related Papers

[1] A. Yaguchi, et al. "Adam Induces Implicit Weight Sparsity in Recti?er Neural Networks," ICMLA 2018 (under review).
[2] K. Isogawa, et al. "Deep Shrinkage Convolutional Neural Network for Adaptive Noise Reduction," IEEE Signal Processing Letters, Vol.25, No.2, 2018.
[3] http://www.toshiba.co.jp/about/press/2017_03/pr1001.htm
[4] http://www.toshiba.co.jp/about/press/2016_10/pr1701.htm

APPLY HERE


Mixed-signal IC design for intelligent power electronics

Explanation

Technology fusion between  microelectronics and artificial intelligence is essential to achieve the ultimate in power electronics, namely, zero-loss, zero-noise, zero-failure, zero-design-cost, and zero-footprint components. Several technologies such as active gate drive and anomaly detection based on device condition monitoring, have been the focus in recent years, led by intelligent power modules (IPMs).
This research focuses not only on the existing themes above, but also explores the new direction of technology fusion to aggressively approach the ultimate power electronics goal.
The fellow’s research will be based on a wide, deep background knowledge and rich ideas, with prior study of areas  including power electronics, IC design, signal processing and/or machine learning. Toshiba provides an excellent environment for such research activities considering the existence of all the related layers of power electronics’ technologies and applications.

Knowledge and Skills Required

Analog/digital IC design skills and/or knowledge on intelligent gate drive/power modules

APPLY HERE


Autonomous behavior acquisition based on media understanding

Explanation

The use of AI-related technologies are widely expected to lead an evolutional change in  value creation and to counter the lack of human workforce in manufacturing, logistics, services and many other commercial fields. In the past, technologies such as autonomous driving based on image recognition and dialogue interface using speech recognition have been actively studied, but going forward, AI technology, which integrates a variety of information and enables autonomous judgment and action, will be required.
The successful candidate will focus on either

  1. Research and development of recognition and understanding of various media such as image, voice and text
  2. Research and development of agents such as next generation autonomous vehicles, robots and drones using automated decision-making and action based on the various information gathered
  3. Research and development of integration of the above to develop a new architecture to make more intelligent decision-making based on accumulated knowledge that can realize a high degree of autonomous activity

Toshiba has many experienced researchers who have a long-standing history of image and speech recognition research, and has many business areas including social infrastructure, electronic devices, and digital solutions  in which to apply the technologies developed.

Knowledge and Skills Required

Candidates are required to have a deep knowledge of at least one of the following fields: pattern recognition, speech recognition, machine learning, knowledge modeling, deep learning, reinforcement learning, and robotics. They should also have software implementation skills such as C, C++, C#, MATLAB(R) and/or python(TM) for building algorithms and conducting experiments.

Related Papers

[1] BiSeg: Simultaneous Instance Segmentation and Semantic Segmentation with Fully Convolutional Networks, Viet-Quoc Pham, Satoshi Ito, Tatsuo Kozakaya, BMVC, 2017
[2] COUNT Forest: CO-Voting Uncertain Number of Targets Using Random Forest for Crowd Density Estimation, Viet-Quoc Pham, Tatsuo Kozakaya, Osamu Yamaguchi, Ryuzo Okada, ICCV, 2015
[3] Random ensemble metrics for object recognition.  Tatsuo Kozakaya, Satoshi Ito, Susumu Kubota, ICCV, 2011
[4] Cat face detection with two heterogeneous features. Tatsuo Kozakaya, Satoshi Ito, Susumu Kubota, Osamu Yamaguchi, ICIP, 2009
[5] Fully automatic feature localization for medical images using a global vector concentration approach, Tatsuo Kozakaya, Tomoyuki Shibata, Tomoyuki Takeguchi, Masahide Nishiura, CVPR, 2008

APPLY HERE


Mechanism studies on Li-ion conduction at the interface between inorganic/organic hybrid semi-solid electrolytes in Li-ion batteries for next generation electric vehicles

Explanation

Toshiba has developed a new type of lithium-ion battery “SCiB™” using lithium titanate anode instead of conventional carbonaceous materials [1]. Compared to other Lithium-ion batteries, the SCiB has remarkably improved safety features plus high-power, long-life and quick-charging, , and has been applied to idling stop systems (ISS), hybrid electric vehicles (HEV), electric vehicles (EV), and energy storage systems (ESS) for power plant. As such, requirements for next generation SCiB batteries have become diversified and not only focused on increased capacity hence our development has been covering a wide field in order to meet these individual requests: high-capacity technologies for EVs, high-power output and high-temperature durability technologies for PHEVs and HEVs, and low cost technologies for Pb-free ESS.
Toshiba has recently developed a hybrid electrolyte, with oxide based solid electrolyte particles coated with small amount of gel polymer electrolyte in order to reduce interfacial contact resistance between solids. A 12V class bipolar battery has been successfully constructed with the hybrid electrolyte, and can operate at freezing point or below [2]. This new technology is important to overcome high internal resistance and/or low chemical stability of solid electrolytes which have been previously reported and to realize high-power and high-energy-density batteries without any liquid electrolytes and sheet separators.
However, the lithium ion conduction mechanism of the hybrid electrolyte is still not fully understood. Not only self-diffusion but also dynamic phenomena in an electric field still needs to be clarified, and this cannot be investigated only with conventional ex-situ experiments. In a complex-modulus-spectroscopy study using an alternating electric field [3], we have identified differences between lithium-ion-conduction phenomena at the surface of hybrid electrolytes and those of conventional electrolytes.Further collaboration with specialists of complex-impedance-spectroscopy in a field of solid state ionics is required in order to obtain a more thorough understanding of the conduction mechanism. We are therefore searching for Fellowship candidates who specialise in solid state ionics and electrochemical analysis. The candidates will be expected to build a design guideline to develop  a more powerful battery with the hybrid electrolyte through the conduction mechanism understanding. 
Toshiba has excellent facilities for materials synthesis, chemical analysis, cell preparation and electrochemical testing including electric furnaces (~1600?), milling machines (wet and dry), dry-rooms, glove-boxes, electrode coating machines, charge-discharge devices and analysis equipment such as complex impedance measuring device (Keysite E4990A), ICP, XRD, SEM, NMR, FT-IR etc..
The successful candidate (Fellow) will therefore be able to experience the entire process from material synthesis to battery testing at our laboratories during the research and gain a wide understanding. While we believe that experience is subservient to knowledge, the goal is to build a strong technology for the Toshiba business in the field of energy infrastructure and electric vehicles. 

Knowledge and Skills Required

Candidates are required to have extensive knowledge in at least one of the following fields: solid state ionics, electrochemistry, computational calculation for ionics, materials chemistry and physics.They should have electrochemical analysis skills such as complex impedance and complex modulus, and materials synthesis skills.

Related Papers
[1] Norio Takami, Hiroki Inagaki, Yoshinao Tatebayashi, Hidesato Saruwatari, Keizoh Honda and Shun Egusa, Journal of Power Sources 244 (2013) 469.
[2] Kazuomi Yoshima, Yasuhiro Harada and Norio Takami, Journal of Power Sources 302 (2016) 283.
[3] I. M. Hodge, M. D. Ingram and A.R. West, Journal of Electroanalytical Chemistry 74 (1976) 125.

APPLY HERE


Artificial intelligence (AI) technologies on the modeling of the relationship between manufacturing process, material factors and battery performance in rechargeable battery production.

Explanation

Toshiba uses simulation to predict performance during the planning and design of its lithium ion rechargeable  battery (LiB). Separately, at the manufacturing stage, we collect both yield data and process control data using IoT technologies.  However, the relationship between battery efficiency and the manufacturing process is complicated and it remains the case that it is difficult to adequately connect them logically. Therefore, we would like to design a system which measures at the earliest stages the relationship between battery performance and the manufacturing process.
At the start of the project to build this system, AI will be applied to data acquired by Toshiba from existing rechargeable battery production, in order to model the relationship among materials, process factors and battery performance. The objective of the project will be to build the methodology for solving the inverse problem, where necessary processes and materials can be predicted based on the battery performance.

Knowledge and Skills Required

  • Knowledge of the manufacturing process and the performance of lithium ion rechargeable batteries
  • Strong interest in data analysis by applying AI, preferably with hands-on experience
  • Strong interest in development of systems to model the whole process using data on lithium ion rechargeable battery production.

Related Papers

[1] T.Kikuchi and M.Ciappa, “Modeling the threshold voltage instability in SiC MOSFETs by multiphonon-assisted tunneling”, Microelectronics Reliability, vol. 58, p.p. 33-38, 2016.
[2] H.Yabuhuhara and A.Miyamoto, “Prediction of low-energy boron profile for ultra-shallow junction formation by hybrid molecular dynamics method”, Japanese Journal of Applied Physics, vol. 55, p. 016503, 2015.

APPLY HERE


Using Machine Learning to Monitor Open Source Software stability, vulnerabilities and compliance information

Explanation

Open Source Software is being used in various products and the number of OSS components in one product can now be in the thousands. As a result, it has become too complex for a human to properly assess the mix of information such as licence and vulnerability of the OSS components. So It has become necessary to reduce the complexity of OSS by automatically collecting and constructing vulnerability and license information.
This research will focus on creating an automated method of collating relevant information including version controls, bill of materials, license checks, security information etc and presenting it to humans in an easy to understand way and to estimate the effectiveness of it.  We expect the successful candidate to propose new methods to  reduce management costs  and the number of errors. We expect the outcome of the research to be applied to real products.

Knowledge and Skills Required

Linux system administration skill, Programming experience (e.g. Java, C, Python), Software version control skill with git et.al, and OSS compliance and vulnerability basic concept.

APPLY HERE


Development of novel compound semiconductor devices by controlling the physical properties of wide-band gap materials

Explanation

Wide-band gap semiconductors with a larger band gap than Si are expected to be applied to next generation power devices and solid-state lighting devices, supporting a future energy-saving society. Although they have outstanding electrical and optical characteristics, their features are not fully utilizable due to the instability and the uncontrollability of their physical properties, because the relation between the crystal defect or the interface state, which are inherent in a wide-band gap semiconductor, and the electronic property has not been clarified. 
The aim of this research is to establish technologies to realise the excellent characteristics of power semiconductor devices with wide-band gap semiconductors, such as, SiC, GaN, AlN, Ga2O3 and to implement the design of novel device structures with the wide-gap semiconductors, or the evaluation, the analysis and the modeling of physical mechanisms.

Knowledge and Skills Required

Candidates should possess expertise in either theory of semiconductor physics or evaluation and analysis of electron devices.

Related Papers

[1] K. Uesugi, et al, "Improvement of Channel Mobility of GaN-MOSFETs With Thermal Treatment for Recess Surface", 'Phys. Status Solidi A 2017, 170051.
[2] T. Yonehara, et al, "Improvement of Positive Bias Temperature Instability Characteristic in GaN MOSFETs by Control of Impurity Density in SiO2 Gate Dielectric'', IEDM17-745.
[3] Y. Kagiwara, et al, "Suppression of Positive Bias Temperature Instability in GaN-MOSFETs", SSDM 2017..

APPLY HERE


Open software stack to operate large numbers of devices

Explanation
Toshiba intends to solve two important problems with the operation of a large number of devices. The word "devices" here refers to computing devices such as smart meters, digital signage, subway gates, air conditioners, or IoT devices for example. The word "operation" refers to the multiple tasks that a "device operator" needs to do to remotely maintain a group of devices. These tasks include status monitoring, rebooting, power control, applying software updates or updating specific settings in those devices.
The first problem that we want to address is caused by the use of closed software. Companies or departments that need to operate devices of a specific type (e.g.: a printer, a POS device, or an air conditioner) often create or order a software stack to operate those particular devices. The end result consists of a non-generic stack that cannot be shared across companies or departments causing duplicated costs. Additionally, creating such systems is not an easy task. For that reason, most software stacks lack functionality and advanced security features. To address this problem, the architecture and software stack created during this Fellowship will be made in collaboration with the open source community.
The second problem is more technical and refers to the increasing amount of devices that operators need to supervise. This large number of devices requires visualization tools and scalable software that is not possible with current approaches. The Fellow will investigate how to address the challenges of operating large amounts of devices.
Part of this Fellowship will consist of investigating open source software stacks written for cloud infrastructure, and evaluating the feasibility of reusing similar concepts to operate large numbers of devices. Unfortunately, cloud infrastructure relies on assumptions that do not always apply to device operation infrastructure. For example, in contrast to cloud virtual machines or containers, end devices are usually placed behind security firewalls or gateways, and therefore direct access from the operator is not always possible. Some of the most popular cloud software such as Ansible or Kubernetes rely on that.
For that reason, the Fellow will need to investigate and adapt concepts from cloud infrastructure to requirements of device operation infrastructure such as indirect or polling-based access.
The researcher will participate in the design of the interfaces and protocols required by each component in the architecture in close contact with the open source community, and will also work on a prototype to validate and evaluate the feasibility of the architecture.

Knowledge and Skills Required

Candidates should have knowledge or deep interest in these topics:

  • IoT systems
  • Distributed software and networks
  • Open source and open collaboration
  • Linux user experience
  • Programming

Related Papers

[1] "How do you update your embedded Linux devices?". LinuxCon Japan, July 14th, 2016. https://events.static.linuxfound.org/sites/events/files/slides/linuxcon-japan-2016-softwre-updates-sangorrin.pdf
[2] "Kernel security hacking for the Internet of Things", LinuxCon Japan, June 3rd, 2015, https://events.static.linuxfound.org/sites/events/files/slides/linuxcon-2015-daniel-sangorrin-final.pdf

APPLY HERE