Ryohei Suzuki

Voice over Sensor Networks

Wireless sensor networks have traditionally focused on low duty-cycle applications where sensor data are reported periodically in the order of seconds or even longer. This is due to typically slow changes in physical variables, the need to keep node costs low and the goal of extending battery lifetime. However, there is a growing need to support real-time streaming of audio and/or low-rate video even in wireless sensor networks for use in emergency situations and short-term intruder detection. In this paper, we present FireFly, a time-synchronized sensor network platform for real-time data streaming across multiple hops. FireFly is composed of several integrated layers including specialized low-cost hardware, a sensor network operating system, a real-time link layer and network scheduling which together provide efficient support for applications with timing constraints. In order to achieve high end-to-end throughput, bounded latency and predictable lifetime, we employ hardware-based time synchronization. Multiple tasks including audio sampling, networking and sensor reading are scheduled using the nano-RK RTOS. We have implemented RT-Link, a TDMA-based link layer protocol for message exchange on well-defined time slots and pipelining along multiple hops. We use this platform to support 2-way audio streaming concurrently with sensing tasks. For interactive voice, we investigate TDMA-based slot scheduling with balanced bi-directional latency while meeting audio timeliness requirements. Finally, we describe our experimental deployment of 42 nodes in a coal mine, and present measurements of the end-to-end throughput, jitter, packet loss and voice quality

Askus: Amplifying Mobile Actions

Information sharing has undeniably become ubiquitous in the Internet age. The global village created on the Internet provides people with instant access to information and news on events occurring in a remote area, including access to video content on websites such as YouTube . Thus, the Internet has helped us overcome barriers to information. However, we cannot conceive an event happening in a remote area and respond to it with relevant actions in a real-time fashion. To overcome this problem, we propose a system called Askus , a mobile platform for supporting networked actions. Askus facilitates an extension of the conceivable space and action by including humans in the loop. In Askus , a persons request is transmitted to a suitable person who will then act in accordance with the request at a remote site. Based on a diary study that led to detailed understanding about mobile assistance needs in everyday life, we developed the Askus platform and implemented the PC-based and mobile phone-based prototypes. We also present the results from our preliminary field trial.

A Protocol for Policy-Based Session Control in Disruption Tolerant Sensor Networks

Recently there has been a great deal of research on using mobility in sensor networks to assist their sensing tasks. In this paper, we propose a policy-based session control protocol for Multi-Robot Sensor Networks (MRSNs) called Billiards. In a MRSN, all messages are transported by the physical motion of participants (mobile nodes) in the network. When a large volume of data or continuous data is required to be transferred, there exists a problem determining how the data is fragmented and how the mobile nodes are formed for carrying the data to the destination. To overcome the issues, we propose a suitable method of session control which is determined based on a state of surrounding mobile nodes such as number, maximum-velocity and buffer-size. Billiards also takes a system policy of delay minimization into consideration. In this paper, we describe the protocol and model of Billiards and analyze the model. We evaluated the performance of Billiards utilizing mobile robots which are equipped with MICA2 mote and comparing with non optimized method. The experimental results demonstrate that Billiards achieves less delay than non optimized method at every velocity and buffer-size of each robot.

An RFID-based human-probe positioning system

We propose an RFID-based human probe positioning system for urban sensing. We call a person who has various sensors on his/her body human probe (HP). Our goals are (1) to estimate a position for HP which is based on location information of RFIDs and other HPs which hold position information with estimated error and (2) to manage and visualize sensor data considering a degree of position accuracy.

Billiards: Policy-based Session Control Protocol in Disruption Tolerant Sensor Networks

Recently there has been a great deal of research on using mobility in sensor networks to assist their sensing tasks. In this paper, we propose a policy-based session control protocol for disruption tolerant sensor networks (DTSNs) called Billiards. In a DTSN, all messages are transported by the physical motion of participants (mobile nodes) in the network. When a large volume of data or continuous data is required to be transferred, there exists a problem determining how the data is fragmented and how the mobile nodes are formed for carrying the data to the destination. To overcome these issues, we propose a suitable method of session control which is determined based on a state of surrounding mobile nodes such as number, maximum-velocity and buffer-size. Billiards also takes a system policy of either delay minimization or energy efficiency into consideration. In this paper, we describe the protocol and model of Billiards and analyze the each model. We also show its effectiveness through experiments with utilizing mobile robots.

Application programming interface for configuration of multi-robot sensor networks

In this paper we propose a design of application programming interfaces (APIs) for configuration of multi-robot sensor networks (MRSNs). Such networks have unique aspects: intentional mobility and delay tolerance. These aspects limit the applicability of conventional approaches proposed for mobile ad-hoc networks (MANETs). APIs for configuration of MRSNs must take these aspects into account. In addition, an application program should be transparent from changing of physical devices such as wireless devices. In this paper we describe the design and prototype implementation of APIs for configuration of MRSNs.

Collaborative 3D Modeling by the Crowd

We propose a collaborative 3D modeling system that deconstructs the complex 3D modeling process into a collection of simple tasks to be executed by nonprofessional crowd workers. Given a 2D image showing a target object, each crowd worker is directed to draw a simple sketch representing an orthographic view of the object, using their visual cognition and real-world knowledge. The system then synthesizes a 3D model by integrating the geometrical information obtained from a collection of gathered sketches. We show a set of algorithms that generates clean line drawings and a 3D model from a collection of incomplete sketches containing a considerable amount of errors and inconsistencies. We also discuss a crowdsourcing workflow that iteratively improves the quality of submitted sketches. It introduces competition between workers using extra rewards based on peer-reviewing as well as an example-sharing mechanism to help workers understand the task requirements and quality standards. The proposed system can produce decent-quality 3D geometries of various objects within a few hours.

CASTIN: a system for comprehensive analysis of cancer-stromal interactome

BackgroundCancer microenvironment plays a vital role in cancer development and progression, and cancer-stromal interactions have been recognized as important targets for cancer therapy. However, identifying relevant and druggable cancer-stromal interactions is challenging due to the lack of quantitative methods to analyze whole cancer-stromal interactome.ResultsWe present CASTIN (CAncer-STromal INteractome analysis), a novel framework for the evaluation of cancer-stromal interactome from RNA-Seq data using cancer xenograft models. For each ligand-receptor interaction which is derived from curated protein-protein interaction database, CASTIN summarizes gene expression profiles of cancer and stroma into three evaluation indices. These indices provide quantitative evaluation and comprehensive visualization of interactome, and thus enable to identify critical cancer-microenvironment interactions, which would be potential drug targets.We applied CASTIN to the dataset of pancreas ductal adenocarcinoma, and successfully characterized the individual cancer in terms of cancer-stromal relationships, and identified both well-known and less-characterized druggable interactions.ConclusionsCASTIN provides comprehensive view of cancer-stromal interactome and is useful to identify critical interactions which may serve as potential drug targets in cancer-microenvironment. CASTIN is available at: http://github.com/tmd-gpat/CASTIN.

Sensorized map using Human Probe

Environment monitoring in urban areas is becoming important to provide information for safety and relief for daily lives. Wireless sensor devices are regarded as means of such monitoring. However, it is not efficient to deploy the sensors more than the necessity in the city. It is efficient to deploy the sensors on the place where many people move around. Therefore, we install sensors to walking people and acquire information to complement the embedded sensors. Based on the idea, we have extended WINFO, human probe, a previously developed system, to accommodate reading peoplepsilas walking states obtained by pressure signals at the peoplepsilas feet. Our goal is to create a map utilizing the information obtained by a collection of the walking signals.

Luigi: Large-scale histopathological image retrieval system using deep texture representations

Background As a large number of digital histopathological images have been accumulated, there is a growing demand of content-based image retrieval (CBIR) in pathology for educational, diagnostic, or research purposes. However, no CBIR systems in digital pathology are publicly available. Results We developed a web application, the Luigi system, which retrieves similar histopathological images from various cancer cases. Using deep texture representations computed with a pre-trained convolutional neural network as an image feature in conjunction with an approximate nearest neighbor search method, the Luigi system provides fast and accurate results for any type of tissue or cell without the need for further training. In addition, users can easily submit query images of an appropriate scale into the Luigi system and view the retrieved results using our smartphone application. The cases stored in the Luigi database are obtained from The Cancer Genome Atlas with rich clinical, pathological, and molecular information. We tested the Luigi system by querying typical cancerous regions from four cancer types, and confirmed successful retrieval of relevant images. Conclusions The Luigi system will help students, pathologists, and researchers easily retrieve histopathological images of various cancers similar to those of the query image.