While the introduction of softwarelization technologies such as SDN and NFV transfers the main focus of network management from hardware to software, the network operators still have to deal with various and numerous network and computing equipment located in network centers. Toward fully automated network management, we believe that a robotic approach will be significant, meaning that robots will handle network-facility management works on behalf of humans. This paper focuses on robotic assistance for on-site network maintenance works. Currently, attendance at the on-site vendor's works (e.g., hardware check, hardware installation/replacement and high impact update of software etc.) is one of the major tasks of network operators. The network operators monitor the work progress for human error prevention and safety improvement. In order to reduce the burden of this, we propose three functionalities, namely attendance at on-site meetings, progress check by periodical patrol, and remote control of robots, which support the various forms of attendance. The paper presents our implementation of enabling these supports, and reports the results of experiments conducted in a commercial network center.

The recent evolution in network technologies such as network function virtualization (NFV) and cloud-native network function (CNF) is expected to enable the delivery of on-demand network services to meet various requirements such as reliability and latency. However, the operation of networks using these emerging technologies can be complex due to the need for a greater number of network entities and more types of service level agreements (SLAs). This incurs a high operational workload. On the other hand, there have been major advances in machine learning (ML) in this decade. These have made it possible for many and diverse industries to leverage ML capabilities to support human tasks. In this paper, we apply ML to analyze the root cause of failures in an NFV environment. To identify the appropriate approach, we analyze the performance of an ML-based fault classifier from several perspectives, three algorithms [multi-layer perceptron (MLP), random forest (RF), support vector machine (SVM)], volume and balance of training data, and number of features. The experiment on the testbed using our ML comparison framework shows that RF provides the highest performance even with a small amount of data, and also that SVM could improve its performance by increasing training data, feature reduction, or balance adjustment of normal/abnormal samples.

While the introduction of the softwarelization technologies such as SDN and NFV transfers main focus of network management from hardware to software, the network operators still have to care for a lot of network and computing equipment located in the network center. Toward fully automated network management, we believe that robotic approach will be significant, meaning that robot will care for the physical equipment on behalf of human. This paper explains our experience and insight gained throughout development of a network management robot. We utilize ROS(Robot Operating System) which is a powerful platform for robot development and secures the ease of development and expandability. Our roadmap of the network management robot is also shown as well as three use cases such as environmental monitoring, operator assistance and autonomous maintenance of the equipment. Finally, the paper briefly explains experimental results conducted in a commercial network center.

Unified builds are a simple but effective technique to reduce the build time of large software projects. Unified builds generate large compiler tasks by bundling multiple source files into one, resulting in a significant reduction in build time through removal of redundant work incurred by shared headers. However, unified builds have a negative effect on incremental builds because each compiler task gets larger. An ad-hoc unification strategy causes an excessive slowdown in incremental builds. A rough report from WebKit says the worst slowdown is 20% (6s → 7s), but our investigation shows it is as high as 479% (19s → 110s).

In this paper, we compared three different types of avatar design (“Body”, “Hand + Arm”, and “Hand only”) for the augmented reality remote instruction system in terms of usability. The result showed that the usability of the remote instruction system with full body avatar has a higher usability. In addition, participants felt more easily to track the full body avatar than the avatar with hand only. However, concerning the understandability of the instruction, there was no difference between three designs.

At times, a composite dataflow needs rerunning in crowdsourcing for various reasons, even when the dataflow may be half complete. Rerunning the dataflow requires more time and incurs monetary costs for the additional work that would need to be completed by crowd workers. This time and cost may be reduced by reusing complete or intermediate results in the previous run. However, at times, such results cannot be used as is (e.g., when the dataflow has been changed), and some additional tasks need to be completed in the old dataflow in order to make them reusable in the new dataflow. The benefit of reusing these results in the previous run may or may not be worth the cost of these additional tasks. This paper gives a general framework for formulating this problem, and proposed a method to estimate the additional costs. The simulation result shows that it is worth devising optimization techniques to identify feasible (namely, cost-effective) plans.

With respect to collaborative physical tasks, gaze and gestures play significant roles when referring to physical objects. In video-mediated communication, however, such nonverbal cues become "ineffectual" when they are presented via a 2D monitor, making video-mediated collaborative physical tasks inefficient. This study focuses on gaze cues to support remote collaborative physical tasks and uses an eye-shaped display, "ThirdEye," a simple add-on display that represents a remote participant's gaze direction. ThirdEye is expected to be especially effective when used with mobile terminals. We investigated whether the ThirdEye's gaze shift is effective in leading a local observer's attention toward objects in the local environment, even when ThirdEye is presented with the actual face image of a remote person. Experimental results show that ThirdEye can lead the local participant's attention to intended objects faster than without ThirdEye.

Graphics Processing Units (GPUs) have become a common computing resource for general-purpose computing (GPGPU). GPU usage has also spread to high-throughput server applications, taking advantage of its massively parallel nature and wide availability at various cloud platforms. Although various methods currently exist to share a single GPU among multiple applications, migrating GPGPU server applications across different machines is challenging due to lack of hardware mechanisms, such as programmable preemption and access to GPU context. This paper presents an event-driven framework for GPGPU server applications, which enables us to implement a software based approach for migration which overcomes current hardware limitations.

In recent years, an increasing number of universities are also using services such as iTunes U, YouTube to make their lectures accessible to the public. In order to improve the quality of video content, we use time stamp and eye tracking together and extract what kind of defect is in the video. A timestamp in this context refers to elapsed time since the start of the video, and it is linked to a viewer. By using this timestamp, the viewers are able to compare their own thoughts to the thoughts of others, thus enabling them to clarify their previously vague ideas in a relatively clear manner. Eye tracking can visualize the potential movement of video viewers and clarify viewers' focus points. We will also construct a platform where this timestamp and eye-tracking can be utilized, and support wide-ranging utilization of video content.

We present novel low-level audio features that are based on correlations between sub-band audio signals decomposed by undecimated wavelet transform. Under the assumption that SVM is used for classifier learning, the experimental results on GTZAN dataset showed that the proposed method demonstrated the best accuracy of 81.5%, outperforming the conventional methods.

A long-standing challenge in video-mediated communication systems is to represent a remote participant's gaze direction in local environments correctly. To address this issue, we developed ThirdEye, an add-on eye-display for a video communication system. This display is made from an artificial ulexite (TV rock) that is cut into a hemispherical shape, enabling light from the bottom surface to be projected onto the hemisphere surface. By drawing an appropriate ellipse on an LCD and placing ThirdEye over it, this system simulates an eyeball. Our experiment proved that an observer could perceive a remote Looker's gaze direction more precisely when the gaze was presented using ThirdEye compared to the case in which the gaze was presented using the Looker's face on a flat display.

Graphics processing units (GPUs) have become an attractive platform for general-purpose computing (GPGPU) in various domains. Making GPUs a time-multiplexing resource is a key to consolidating GPGPU applications (apps) in multi-tenant cloud platforms. However, advanced GPGPU apps pose a new challenge for consolidation. Such highly functional GPGPU apps, referred to as GPU eaters, can easily monopolize a shared GPU and starve collocated GPGPU apps. This paper presents GLoop, which is a software runtime that enables us to consolidate GPGPU apps including GPU eaters. GLoop offers an event-driven programming model, which allows GLoop-based apps to inherit the GPU eaters' high functionality while proportionally scheduling them on a shared GPU in an isolated manner. We implemented a prototype of GLoop and ported eight GPU eaters on it. The experimental results demonstrate that our prototype successfully schedules the consolidated GPGPU apps on the basis of its scheduling policy and isolates resources among them.

In this study, we used a humanoid robot as a telepresence robot and compared with the basic telepresence robot which can only rotate the display in order to reveal the effect of embodiment. We also investigated the effect caused by changing the body size of the humanoid robot by using two different size of robots. Our experimental results revealed that the embodiment increases the remote person's social telepresence, familiarity, and directivity. The comparison between small and big humanoid robots showed no difference and both of them were effective.

Graphic processing units (GPUs) provide a massively-parallel computational power and encourage the use of general-purpose computing on GPUs (GPGPU). The distinguished design ofdiscrete GPUshelps them to provide the high throughput, scalability, and energy efficiency needed for GPGPU applications. Despite the previous study on GPU virtualization, the tradeoffs between the virtualization approaches remain unclear, because of a lack of designs for or quantitative evaluations of the hypervisor-level virtualization for discrete GPUs. Shedding light on these tradeoffs and the technical requirements for the hypervisor-level virtualization would facilitate the development of an appropriate GPU virtualization solution.$\sf{GPUvm}$, which is an open architecture for hypervisor-level GPU virtualization with a particular emphasis on using the Xen hypervisor, is presented in this paper.$\sf{GPUvm}$offers three virtualization modes: the full-, naive para-, and high-performance para-virtualization.$\sf{GPUvm}$exposes low- and high-level interfaces such as memory-mapped I/O and DRM APIs to the guest virtual machines (VMs). Our experiments using a relevant commodity GPU showed that$\sf{GPUvm}$incurs different overheads as the level of the exposed interfaces is changed. The results also showed that a coarse-grained fairness on the GPU among multiple VMs can be achieved using GPU scheduling.

A long-standing challenge of video-mediated communication systems is to correctly represent remote participant gaze direction in local environments. A telepresence robot with a movable display that shows the face of a remote participant is a promising approach for solving this issue. Researchers generally consider that display orientation is effective for local participants to properly estimate the gaze direction of remote participants. We investigate how subjects estimate gaze direction of a remote participant ("Looker") when his/her face is displayed on a rotatable flat display. Our experiment reveals that both the Looker's head-eye rotation in the display and display rotation affect subject estimation, but the effect of the display rotation is relatively small. Furthermore, we reveal that subjects tend to overestimate Looker gaze direction. Based on our results, we propose a design implication for a telepresence robot to reduce overestimation and properly represent the remote participant gaze direction.

To realize a new type of personal vehicle for elderly people while contributing to a low-carbon traffic society, we have been developing a robotic wheelchair with autonomous traveling and obstacle avoidance in an urban environment. In this paper, we primarily discuss two kinds of key technologies, long-distance autonomous travel in an outdoor environment and obstacle avoidance in a human-shared environment. For the localization and navigation methods, we propose sub-map dividing and realignment with FastSLAM, which enables generation of large-scale 3D voxel maps by the sampling based SLAM method. For the planning and obstacle avoidance methods, we propose motion control by a combination of the global/local A* algorithm and the dynamic window approach. To confirm the effectiveness of our proposed methods, the results of demonstration experiments using our robotic wheelchair ‘Marcus’ in the Tsukuba Robot Zone are also reported.

In this paper, we examine the re-embeddability of maximum 1-planar graphs. In particular, we prove that every optimal 1-planar graph is uniquely 1-embeddable on the sphere except for a sequence of graphs that are minimal with respect to certain reductions. These optimal 1-planar graphs are closely related to their quadrangular subgraphs. We also give a generating theorem for optimal 1-planar graphs.

In the recent situation that robotics is expected for education of technologies and sciences in Japan, it is important to consider what psychological states should be achieved in students, and understand how far the current states are from the states to be achieved, especially from the perspective of ldquomaking-artifactsrdquo in the sense of manufacturing industries. This paper reports some case studies to explore psychological relationships between robots, ldquomaking-artifacts,rdquo computers, and mathematics in younger ages. Then, the paper discusses implications on the practices of robotics in education of technologies and sciences.

This paper presents and describes the approach for achieving long distance autonomous navigation with a mobile robot on outdoor cluttered pedestrian paths. The task was to finish an event launched by the City of Tsukuba in Japan, called “RealWorld Robot Challenge”, of navigating 1km autonomously in a real environment with real pedestrians and bicycles. The hardware, software and strategy for navigating in cluttered environments is explained. Moreover, the complementary functionality of the overall system where map-based and sensor-based navigation seamlessly change, is presented. The robustness of the system is validated with experimental results.