論文

個別に設計された制御器が相互作用しあう分散システム全体の振る舞いや安定性について定式化・解析し、その上で複数制御器の構成法および制御器設計の手続きを含む総合的な分散システムの制御系設計手法を提案している。

Processes of network-based control system require accurate timing control of frame transmission. A special network interface card (NIC) was proposed and implemented to adjust the timing accurately based on a dual queue structure and precise time synchronization mechanism in this paper. A high resolution measurement tool for frame transmission latency was also developed to validate the implementation. The measured results verified that the proposed NIC achieved the latency shorter than 790 ns and the deviation of it less than 49 ns while a commercial off-the-shelf NIC showed the latency longer than 9700 ns and the deviation of it more than 300 ns.

This paper proposed a precise time synchronization method between schedulers of real-time operating systems by communicating time stamps over a network. An NTP-based communication protocol was used to measure the time offset between nodes. To minimize the measurement error, a light-weight network stack called RT-Messenger was used instead of common transmission control protocol/Internet protocol (TCP/IP) stack. Also, a Kalman filter was introduced to eliminate such error and to estimate time offset and drift. Furthermore, a real-time scheduler in RTAI/Linux (where RTAI is real-time application interface) was modified to compensate the time difference using the estimated offset and drift. The time compensation of a scheduler clock affects the next resume times for real-time tasks. Therefore, a recalculation process of next resume times for all real-time tasks and a reordering process of real-time task queue were added to the real-time scheduler. To validate the proposed method, time differences of resume times for two real-time tasks on different nodes were measured under various experimental conditions. The results confirm that the resume time differences were successfully kept lower than 400 ns with 300 ns standard deviation under the condition of observation period at 1 s and correction period at 10 s.

This paper proposed a precise time synchronization method between schedulers of real-time operating systems by communicating timestamps over a network. An NTP based communication protocol was used to measure the time offset between nodes. To minimize the measurement error, a light-weight network stack called RT-Messenger was used instead of common TCP/IP stack. Also, a Kalman filter was introduced to eliminate such error and to estimate time offset and drift. Furthermore, a real-time scheduler in RTAI/Linux was modified to compensate the time difference using the estimated offset and drift. The time compensation of a scheduler clock affects the next resume times for realtime tasks. Therefore, a re-calculation process of next resume times for all real-time tasks and a re-ordering process of reak-time task queue were added into the real-time scheduler. To validate the proposed method, time differences of resume times for two real-time tasks on different nodes were measured under various experimental conditions. The results confirm that the resume time differences were successfully kept lower than 400 ns with 300 ns standard deviation under the condition of observation period at 1 s and correction period at 10 s.

Predicting the human visual attention in an image is called saliency prediction and is an active research area in the field of neuroscience and computer vision. Early works on saliency prediction was performed by using low-level features. In recent years, convolutional neural networks have been adapted for saliency prediction and achieved the state-of-the-art performance. However, the eye-gaze depends on the personality of each viewer and conventional methods did not take into account such individual properties of the viewer. Therefore, this paper proposes a novel saliency prediction method considering the influence of eye-gaze. Assuming that personality can be expressed as the degree of attention to an object, our proposed method considers the personality by learning which objects are likely to be perceived by each viewer and weighting the universal saliency map with the generated mask based on the object detection results. The experimental results show that the proposed universal saliency map achieves higher accuracy than conventional methods on the public dataset, and the proposed weighted saliency map can reflect the variation of the eye-gaze influences among viewers.

Predicting the human visual attention in an image is called saliency prediction, and is an active research area in the field of neuroscience and computer vision. Early works on saliency prediction was performed by using low-level features. In recent years, convolutional neural networks (CNN) have been adapted for saliency prediction and achieved the state-of-the-art performance. However, the eye-gaze depends on the personality of each viewer and conventional methods did not take into account such individual properties of the viewer. Therefore, this paper proposes a novel saliency prediction method considering the influence of eye-gaze. Assuming that personality can be expressed as the degree of attention to an object, our proposed method considers the personality by learning which objects are likely to be perceived by each viewer, and weighting the universal saliency map with the generated mask based on the object detection results. The experimental results show that the proposed universal saliency map achieves higher accuracy than conventional methods on the public dataset, and the proposed weighted saliency map can reflect the variation of the eye-gaze influences among viewers. (1)

M2M (machine-to-machine) is required and spreads rapidly in recent years. Wireless multi-hop communication is one of the communication styles to construct M2M. However, the performance estimation of such wireless multi-hop communication is not so easy because the bit error ratio and symbol rate adaptation heavily depend on various environmental conditions. Stochastic network calculus (SNC) is a theory for the performance estimation of network. However, current SNC lacks ways to take wireless random errors, branching and joining of traffic flows into calculation, these are necessary to estimate the performance of sink-tree wireless-sensor network. This paper proposed a performance estimation method for sink-tree wireless multi-hop network based on SNC. The proposed method has three attractive points. Firstly, this introduces parameters of hop number and node number considering wireless random error due to packet collision. Secondly, this treats actual sink-tree network model by considering joining of traffic. Thirdly, this separates the service curve and bounding function calculation to avoid double counting. To evaluate the accuracy of the result of the proposed method, conventional estimation methods and simulation evaluation were conducted on a typical wireless multi-hop sensor data collection network. The results validate that the proposed method achieves much higher-accuracy than conventional methods.