論文

Shape changing robot is an approach towards locomotion on uncertain terrain due to its omni-directional features. However, the current locomotion method for such robots rely on discontinuous rolling. We propose a free form locomotion: an omni directional continuous crawling for deformable robots. This method introduce continuous shifting of contact surface similar to amoeba movement. A Mochibot that has thirty two telescopic legs is developed to verify the proposed locomotion method. Through the experiments, we have confirmed that the robot can track smooth paths: straight, smooth, and hand written curves. We also evaluate errors between desired and measured trajectories of the robot.

Due to the advance in information and communication technologies, smart cities have become an increasingly appealing vision. One indispensable basis in the research, development and implementation of smart cities is urban sensing technology. Automotive sensing is a novel urban sensing technology that equips sensors to vehicles, like buses, taxi cabs or garbage trucks and utilizes their mobility to help to conduct urban sensing tasks. In this paper, we present Cruisers, an automotive sensing platform for smart cities, based on the following ideas: a) Garbage collecting trucks are used to host sensors, b) Cellular communication network is used to deliver sensed data to servers, and c) Proxy server(s) are adopted to convert the format of sensed data to required ones, such as sampling rate, spatial interval or pub/sub protocol. To validate the feasibility of Cruisers, we implement a real system that consists of 66 customized sensing modulars integrating typical environmental and mobility sensors and deploy them into the garbage collecting trucks at Fujisawa, a medium-sized city in Japan. Performance evaluation is conducted to investigate the performance of Cruisers in terms of scalability, data publishing delay and spatial coverage.

We propose a damaged traffic lane detection method ensuring high accuracy in spite of only a few number of supervised data which are labeled traffic lane images. In general, supervised machine learning approach is very powerful for image classification. However, preparing a large amount of supervised data is time-consuming task because labeling damaged or not damaged is usually done manually through visual inspection of images. Thus, lowering the cost of labeling data is a great concern. To this end, we adopt a semi-supervised machine learning approach which learns from both labeled and unlabeled data by constructing graph based on the image similarity. We captured a large amount of the road lane images. Then, we constructed graph structure whose nodes are the road lane images and whose edges are the similarity between the images. In several nodes, we assigned labels which denote "damaged" or "not damaged." Finally we utilized the label propagation, which made it possible to infer the labels of the unlabeled data from the labeled data. These estimation resulted in the accuracy rate over 85% from the supervised data, which accounted for only 1.8% of the total data.

This paper proposes DeepCounter, an automotive sensing system where deep learning based image processing technology is used to automatically count the number of collected garbage bags from the video taken by a camera mounted on the rear of a garbage truck in order to sense a fine-grain spatio-temporal distribution on the amount of disposed garbage in cities that is envisioned to be helpful to develop novel applications related to garbage collection there. A prototype system is implemented on a GPU-integrated signal-board computer. A detection-tracking-counting (DTC) algorithm is developed and implemented based on the single shot multibox detector (SSD), a well-known real-time object detection algorithm. Experimental evaluation validates the feasibility of the proposed approach using video of realistic garbage collection in Fujisawa city, Japan.

We propose a Meal Photo SNS (called HealthyStadium) for improving users' eating habits by mutually assessing each others' health. This application's method is to evaluates pictures of meals leads to the realization of sustainable journaling. In addition, we implement competitive awareness that motivates the users to improve their eating habits by allowing users to share their healthiness ranking amongst the users. Through our study, we confirmed that our system enables users to motivate for improving eating habits. Also we found the number of records is increased by our system.

In correlation with the socio-economic development, changes in people’s lifestyle brought about significant impact on dietary patterns. Though public concerns over healthy eating are increasing, many are still uncertain when choosing a well balanced meal amid welter of information. In this paper, we propose “KomaFLens”, a mobile system and application built for Microsoft HoloLens, which aims to motivate long term dietary habit modification through gamification. The primary purpose of this research is to enhance the users’ nutritional knowledge and to guide them to make healthier choices in their diet. Our preliminary evaluation revealed interesting points for discussion regarding the procedure for capturing food labels. Streamlining the operational method to boost tractability will improve the accuracy when recording food intakes.

Chemotherapy-induced peripheral neuropathy (CIPN) is a common side effect of anticancer drugs that causes muscle weakness in the cancer patients, causing them to fall. Therefore, we constructed "FD-AWARE", a system to understand the users' fall context and users' CIPN symptoms as the first step in preventing these falls. This system can collect the various sensor data from the iPhone and the Apple Watch, self-reported fall information data, self-reported user status data of CIPN symptoms, and their physical condition. We conducted a 2-week in-the-wild experiment with 8 patients who were actually suffering from CIPN. We constructed the machine learning models for estimating the users' status of CIPN symptoms and successfully achieved high accuracy of performance for several estimating models.

Sharing and analyzing image data from ubiquitous urban cameras must enable us to understand and predict various contexts of the city. Meanwhile, since such image data always contains privacy data such as people and cars, we cannot easily share and analyze the data through the Internet for the viewpoint of privacy protection. As a result, most of urban image data are only kept/shared within the camera owners, or even discarded to reduce risks of privacy data leakage. To solve the privacy problem and accelerate sharing of urban image data, we propose GANonymizer that automatically detects and removes objects related to privacy from the urban images. GANonymizer combines two neural networks: (1) a network which detects objects related to privacy such as persons and cars in an input image using object detection network and (2) a network that removes the detected objects naturally as though they are not exist originally. Through our experiment of applying GANonymizer to urban video images, we confirmed that GANonymizer partially achieved natural removal of objects related to privacy.

Happiness is obviously one of the most fundamental essence that affects many aspects of our lives. Past research found that happiness of one person affects that of other people. What occurs under this propagation of emotion is called “emotional contagion,” a phenomenon wherein through perception, people experience the same emotion expressed by someone when communicating with them. Although online communication is increasing due to growth of mobile computing, emotional contagion on online communication is not well studied yet. Particularly, it is not yet clear if emotional contagion among people occurs through selfie photographs posted on the social network media. We implemented “SmileWave,” the social networking system for investigating selfie-based emotional contagion. The key feature of SmileWave is detecting “smile degree” in user’s posting selfies and in reactive facial expressions when the user is viewing the posted photographs from others. Our in-the-wild user studies with 38 participants for 2 weeks revealed the occurrence of selfie-based emotional contagion over the social network, based on the results that the users’ smile degree improved (15% on average) when the user looked at posted selfie photographs.

The latest smartphones have advanced sensors that allow us to recognize human and environmental contexts. They operate primarily on Android and iOS, and can be used as sensing platforms for research in various fields owing to their ubiquity in society. Mobile sensing frameworks help to manage these sensors easily. However, Android and iOS are constructed following different policies, requiring developers and researchers to consider framework differences during research planning, application development, and data collection phases to ensure sustainable data collection. In particular, iOS imposes strict regulations on background data collection and application distribution. In this study, we design, implement, and evaluate a mobile sensing framework for iOS, namely AWARE-iOS, which is an iOS version of the AWARE Framework. Our performance evaluations and case studies measured over a duration of 288 h on four types of devices, show the risks of continuous data collection in the background and explore optimal practical sensor settings for improved data collection. Based on these results, we develop guidelines for sustainable data collection on iOS.

We propose in this paper an outdoor positioning system based on Wi-SUN FAN network, the goal of which is to protect the elderly, young children and even pets via estimating their locations in a city. In order to achieve long-term portability and network-side positioning, the system does not directly rely on GPS receiver mounted on terminals but use machine learning for location estimation via the received signal strength indication (RSSI) measurements. In particular, the system consists of Wi-SUN beacons, Wi-SUN base-stations and vehicular devices. A beacon, attached to the one to be positioned, broadcasts wireless signal periodically so that its location can be estimated using machine learning algorithms from the RSSIs measured at multiple base-stations that are densely deployed over a city to construct an ad hoc network. Using the mobility of vehicles that roam over a city routinely, such like garbage collection trucks, buses and taxies. Vehicular devices containing both a Wi-SUN beacon and a GPS are used to collect RSSIs and the corresponding GPS coordinates to train the estimation models. We develop a prototype system consisting of 9 base-stations and deploy it to our university campus to conduct a field experiment to validate the proposed approach. Offline analysis on the data collected from the experiment showed that a RandomForest learner performs best among four selected learning algorithms using the default parameters of Weka 3.8, which achieves a mean absolute error of 35.43m and a root mean squared error of 44.21m, respectively. Evaluation on network performance is also conducted.

In this paper, we demonstrate the multiple points of innovation when combining blockchain technology with Internet of Things (IoT) and security frameworks. The deployment and use of IoT device networks in smart city environments has produced an enormous amount of data. The fact that those data are possessed by multiple sources that use independent systems for data collection, storage, and use impedes the exploitation of their value. Blockchains, as distributed ledgers, can be used for addressing the development of a universal system for data collection and distribution. Smart contracts can be used to automate all the processes of such a network, while at the same time, blockchain and the InterPlanetary File System (IPFS) protect sensitive data through anonymity and distributed storage. An innovative and open IoT blockchain market of applications, data, and services is proposed that: (i) provides the framework upon which objects and people can exchange value in form of virtual currencies, for assets (data and services) received; (ii) defines the motivation incentives according to social and business context for humans and smart objects to interact. The specific marketplace is piloted through a cross-border trial between Santander and Fujisawa, in the context of the M-Sec project, validating thus the interoperability, efficiency, and data protection principles.

In this study, we construct an object detection model that targets billboard advertisements in images and detects them regardless of their content and models to extract multiple elements from billboards. In addition, we try multiple methods for each task to clarify an appropriate method. As a demonstration, we prepare a web application that extracts elements from images of billboards by using the model constructed in this study.

With an increase in the number of IoT devices, the amount of data transferred between the devices and applications is becoming huge. To ensure that these data are properly used, a new IoT data transfer system is needed to better control the timing and the content of the data to transmit. One of the promising means for such a large-scale city-data transfer is the publish/subscribe messaging model, which can separate data senders and receivers so that they can run independently. However, existing pub/sub systems cannot cope well with the mobility of senders and receivers, thereby limiting its applicability to real-world uses. In concrete, they don't consider Time-to-live (TTL) of data. Users can use the data anytime and within TTL of it. IoT platforms can improve controlling data transmission by used to this characteristic. In this paper, we focus on the Time-to-Live of data (data-TTL). Our system can control data transmission by using data-TTL and combine with the user's movement information. We have constructed a system, that is capable of control data transmission for mobility aware.

Aquarium fish is one of the most common household companion animals for humans. Keeping and appreciating aquarium fish enriches us and provides comfort to our lives. However, difficulty in communicating and understanding their fish may negatively impact the person's motivation to continuously take care of their fish. In order to enhance human-fish interaction, we propose “AffectiveNemo”, an ACI system for aquarium fish, that estimates and displays “observed emotion” (defined as fish's emotional states that the system estimates based on observations by humans), via speech bubbles shown on an LCD display placed behind the aquarium. The system estimates the observed emotion by analyzing the movement of the fish in the aquarium, and combining this data along with the emotions that the humans perceive the fish to have. We evaluated the accuracy of each module as a preliminary experiment, conducted a user study on the influence of displaying observed emotion, and evaluated the utility of the system. Our findings concluded that the system can correctly detect the observed emotion, further enhance viewer sensitivity by displaying the observed emotion of aquarium fish that humans are perceptive, and promote the interest in aquarium fish.

To achieve behavior change in our lives for several specific goals towards better well-being, various types intervention systems have been proposed. However, because of humans' bounded rationality, such explicit information provision for the users do not always reach their attention, thus fails the needed information for the behavior change. This paper propose new concepts "behavification"with which information system bypasses the user's cognitive, attentional and perceptive system by not notifying him/her and modifies the system behavior to directly influence the user's behavior under user's explicit advance permission. Through a series of group discussion and survey sessions with 19 participants, we investigated how the potential behavification users see its appropriateness to different application categories, killer application scenarios, and design guidelines and rationales for a behavification system to support such scenarios. We found total 58 application scenarios and propose 3 guidelines "advance user permission", "multi-modality"and "integration with real-world action/actuation".