コロキアムB発表

日時: 9月27日(木)1限(9:20~10:50)


会場: L1

司会: 新谷 道広
上竹 規之 M, 2回目発表 コンピューティング・アーキテクチャ 中島 康彦, 井上 美智子, 中田 尚, TRAN THI HONG, 張 任遠
title: Programmable Analog Calculation Unit for Vector Computations
abstract: Programmable analog calculation unit is designed for computing arbitrary functions of highly dimensional vectors in real-time. The support vector regression algorithm with Gaussian kernels is employed to approximately retrieve vector calculations. Through the proposed efficient scheme of regression, the number of support vectors is greatly reduced. A compact analog circuit is implemented to generate dynamically tunable Gaussian kernel functions of vectors. By integrating the kernel generation circuits in fully parallel, arbitrary vector functions are retrieved without any clock-cycle latency. The proof-of-concept processor is designed in a standard 0.18 um CMOS technology for vector calculations with the maximum dimensions of nine, which is implemented by only 9460 transistors.
language of the presentation:Japanese

 
菊谷 雄真 M, 2回目発表 コンピューティング・アーキテクチャ 中島 康彦, 井上 美智子, 中田 尚, TRAN THI HONG, 張 任遠
title: Development and Implementation of a Liner Array Accelerator with Multithreading
abstract: The benefits of low power consumption and cost reduction by miniaturization of semiconductors are decreasing. Therefore, a paradigm shift towards computing infrastructures prioritizing low power and easy-to-estimate performance is arising at the expense of programmability. 2D-array architecture is one of candidates for such types of still-programmable accelerators. However, the architecture with sufficient local memory and arithmetic resources often brings a disadvantage in performance per area due to the enormous number of wires connecting many memory blocks and ALUs. In our study, in order to improve the area efficiency of 2D-arrays, we propose a multithreading mechanism that reduces the number of physical columns using a ring structure. In this presentation, we report evaluation results and comparison with our conventional 2D-array accelerator by using FPGA prototype system.
language of the presentation: Japanese
 
平賀 由利亜 M, 2回目発表 コンピューティング・アーキテクチャ 中島 康彦, 井上 美智子, 中田 尚, TRAN THI HONG, 張 任遠
title: High efficient Multi-task neural network with shared CNN
abstract: Modern deep learning has significantly improved the performance and has been used in a variety of applications. Due to the heavy processing cost, major platforms for deep learning have been migrated from commodity computers to the cloud where have huge amount of resources. However, the above situation leads to the slowdown of response time due to severe congestion of the network traffic. To alleviate the overconcentration of data traffic and power consumption, many researchers have paid attention to edge computing. I tackle with the parallel processing model using Deep Convolutional Neural Network (DCNN) employed on multiple devices, and the size reduction of network traffic among the devices. I propose a technique that compresses the intermediate data and aggregates common computation in DCNN for multiple tasks.
language of the presentation: Japanese
 

会場: L2

司会: 崔 允端
LIU CHENGBO D, 中間発表 ネットワークシステム学 岡田 実, 杉本 謙二, 東野 武史, Duong Quang Thang

Title: Reliable and efficient wireless communication in narrow space

Abstract: With IoT development, more and more devices will be deployed in narrow space such as factory to increase the productivity. However, it is difficult to maintain the reliable communication environment due to interference and Doppler shift. On the other hand, it is necessary to increase the bandwidth efficiency (BWE) due to frequency resources limitation. OFDM is widely utilized for many applications such as wireless LAN. We combined OFDM system with subcarrier index modulation (SIM) and proposed two systems which are called half-symbol BPSK modulated SIM-OFDM (HS/BPSK/SIM-OFDM) and double ASK-SIM for OFDM system to increase the bandwidth efficiency. On the other hand, the channel state information (CSI) such as RSSI prediction can be utilized to adjust the devices to improve the reliability.

 

Language of the presentation: English

 

 

 

 
佐久間 達也 M, 2回目発表 ロボティクス 小笠原 司, 杉本 謙二, 高松 淳, 丁 明, Gustavo Garcia
title: A Novel Grasping Mechanism for Wide Variety of Objects Using Optical Sensing and Jamming Transition
abstract: In order for a robotic gripper to perform as well as the human hand, the gripper's tactile sensing performance and grasping performance must be improved. We propose a novel grasping mechanism combining both object holding using the jamming transition and optical sensing function. By using a transparent material along with the jamming transition, we both achieve a tactile sense via optical sensing and achieve improvement in the holding stability by tightly restraining the object. In addition, because the hardness of the membrane's contact region can be adjusted, the gripper can grasp even fragile or heavy objects.
language of the presentation: Japanese
 
土屋 裕杜 M, 2回目発表 ロボティクス 小笠原 司, 杉本 謙二, 高松 淳, 丁 明, Gustavo Garcia
title: Pouring motion from deformable container based on tactile information using dual arm robot
abstract: Many household tasks involve pouring operations. Previous research, pouring by a robot is premised on pouring from bottles and cups. However, doesn't consider of deformable containers that deform during operation of such as pouches, plastic bags In this presentation, we proposed a pouring operation from a deformable cotainer by a dual arm robot equipped with a force sensor at the fingertip.The method can reduce the shape of the deformable container due to the wide contact area and grasp with small gripping force. We measure shape change of an object by motion capture system and compare single arm and dual arm.
language of the presentation: Japanese
 
安田 晃也 M, 2回目発表 ロボティクス 小笠原 司, 杉本 謙二, 高松 淳, 丁 明, Gustavo Garcia
title: Packing planning with considering constraints from loading various type object for robotic manipulation
abstract: There are many tasks of placing and loading objects in limited areas such as boxes and bags. In this research, we aim to automate object placement in limited area by robot. In the planning system that optimizes the current loading, there is no calculation involving a variety of constraints that occur when carrying loads of mixed solids and soft objects. Therefore, we extend the evaluation of whether the constraint is satisfied from the position relation of the object to the evaluation function which performs the optimization, and obtain the stable arrangement order from the feature and position of the object after placement. Based on the calculation results, we aim to pack of various objects on cardboard using robots.
language of the presentation: Japanese