Low power dissipation and maximum battery run-time are crucial in portable electronics and EV’s. Battery characteristics and performance varied at different operating conditions. By using accurate, efficient circuit and battery models, designers can predict and optimize battery runtime, current state of charge (SOC) and circuit performance. A great factor in determining the stability of battery system lies within the state of charge estimation. Failing to predict SOC will cause overcharge or over discharge which potentially will bring permanent damage to the battery cells. Open circuit voltage (OCV) has been widely used to estimate the state of charge in estimation algorithms. This paper proposed an accurate and comprehensive battery state of charge (SOC) estimation method by using the Kalman filter. First, Kalman filter for Li-ion battery state of charge estimation was mathematically designed. Then Electrical battery model is being implemented with Kalman filter in matlab Simulink to estimate the exact battery state of charge using estimated battery open circuit voltages. The proposed model shows that system is estimating battery state of charge more accurately than commonly used methods which can help to improve battery performance and lifetime.
This paper proposes a personal-value based item modeling, which is used for calculating predicted ratings and for explaining recommendation. Personal value is one of factors affecting our decision making, and its application to recommender systems has been studied recently. This paper extends existing personal values-based user modeling to item modeling, which estimates characteristics of reviewers who like / dislike target items. A method for calculating predicted ratings based on obtained personal values-based item models is also proposed. Furthermore, this paper focuses on explanation of recommendation as well, which is one of challenges in the recent study of recommender systems. Improvements of user’s satisfactions for recommender systems by showing process of recommendation gets to be important in addition to precision of recommendation. A recommender system is developed based on the proposed method, of which effectiveness is evaluated by a user experiment, in which the target items are movies. Experimental results showed the effectiveness of the proposed method including recommendation accuracy and an explanation of recommendation. It is also shown that the proposed recommender system has the potential to recommend long-tail items.
This paper proposes a method for classifying street lighting conditions after dark in order to share the collected data with the local community. Such information is important for the safety and security of residents, and can be used to discuss about anti-crime activities and nighttime route recommendations. However, it is difficult to ascertain the actual street lighting conditions because of insufficient street-lamp data and the effects of obstacles and other light sources. In order to tackle this problem, we propose a social approach by which local residents collaboratively collect street lighting conditions using their smartphones. The technology behind this approach is a classifier that places the street lighting conditions into one of three levels. It is based on three attributes that are calculated from the illuminance data collected by the smartphones. The results of experiments on 164 actual streets show a maximum classification accuracy of 88.4%. We also discuss performance differences between smartphones and the effect of walking speed during data collection, both of which are important factors affecting the classification accuracy.
A point of interest (POI) is a specific location that people may find useful or interesting, such as restaurants, stores, attractions, and hotels. With the recent proliferation of location-based social networks (LBSN), numerous users gather to interact and share information on various POIs. POI recommendations have become a crucial issue because it not only helps users to learn about new places but also gives LBSN providers chances to post POI advertisements. As we utilize a heterogeneous information network to represent an LBSN in this work, POI recommendations are remodeled as a link prediction problem, which is significant in the field of social network analysis. Moreover, we propose to utilize the meta-path-based approach to extract implicit but potentially useful relationships between a user and a POI. Then, the extracted topological features are used to construct a prediction model with appropriate data classification techniques. In our experiments, the Yelp dataset is utilized as our testbed for performance evaluation purposes. The results show that our prediction model is of good prediction quality in practical applications.
Mongolian and Chinese statistical machine translation (SMT) system has its limitation because of the complex Mongolian morphology, scarce resource of parallel corpus and the significant syntax differences. To address these problems, we propose a template-based machine translation (TBMT) system and combine it with the SMT system to achieve a better translation performance. The TBMT model we proposed includes a template extraction model and a template translation model. In the template extraction model, we present a novel method of aligning and abstracting static words from bilingual parallel corpus to extract templates automatically. In the template translation model, our specially designed method of filtering out the low quality matches can enhance the translation performance. Moreover, we apply lemmatization and Latinization to address data sparsity and do the fuzzy match. Experimentally, the coverage of TBMT system is over 50%. The combined SMT system translates all the other uncovered source sentences. The TBMT system outperforms the baselines of phrase-based and hierarchical phrase-based SMT systems for +3.08 and +1.40 BLEU points. The combined system of TBMT and SMT systems also performs better than the baselines of +2.49 and +0.81 BLEU points.
Applying brain signals to human-computer interaction enables us to detect the attention. Based on P300 signals – one type of event-related potential – enables brain-machine interface users to select desired letters by means of attention alone. Previous studies have reported the feasibility of P300 signals in enabling a single subject to realize novel information retrieval. In the recent collaborative EEG study of multiple subjects has enabled classification to detect attention in a markedly improved way. Here we propose emotional face retrieval using P300 signals of 20 subjects. As a result, the F-measure under the condition of a single subject was a standard deviation of 0.636 ± 0.05 and an F-measure of 0.886 with multiple subjects. In short, emotional face retrieval classification is improved with collaborative P300 signals from multiple subjects. This technique could be applied to life logs, computer-supported cooperative work, and neuromarketing.
The context search engine has been studied for answering trend-related queries. As trend information is obtained from temporal data, which is common in many applications, the context engine is expected to be available regardless of domains. When using existing search engines, it is supposed that users submit a series of queries based on search intention. Therefore, search functions of the context search engine should be designed based on the user’s potential search intention. To analyze user’s behavior in information retrieval, this paper conducted experiments using existing Web search engines. The experimental result is analyzed, based on which the design of a context search engine is described. As another contribution of this paper, new types of temporal variations which can be used to specify queries of the context search engine are also proposed. The results of user experiments confirmed the usability of the proposed temporal variations.
Cognitive sharing of objects is fundamental in a heterogeneous robot system composed of a Unmanned Aerial Vehicle and a ground robot. Since the viewpoint of a UAV is greatly different from a ground robot, they may have different perceptions about the same objects. That makes it difficult to realize cognitive sharing. In this paper, we proposed a cognitive sharing method which is based on Geometric Relation-based Triangle Representations. The method is able to make a UAV and a ground robot identify the same object from similar objects without sharing appearance information in unstructured environment. To copy with the problem of increasing computational cost for the recognition of objects in the Region of Interest, entropy evaluation is employed to evaluate and select unique representations. We illustrated the proposed method with robots in real world.
HBase and Cassandra are two most commonly used large-scale distributed NoSQL database management systems; especially applicable to a large amount of data processing. Regarding remote data backup, each kind of datacenter has its own backup strategy to prevent the risks of data loss. With Thrift Java, this paper aims to implement in-cloud high efficient remote datacenter backup applied to in-cloud NoSQL databases like HBase and Cassandra. The binary communications protocol technology from Apache Thrift is employed to establish the graphical user interface instead of the command line interface so as to ease data manipulation. In order to control the network traffic flow smoothly, intelligent adaptation using ANFIS and PSO is employed to tune the parameters of NoSQL databases during the remote data backup to improve QoS in the network. The stress test has taken on strictly data reading/writing and remote backup of a huge amount of data to verify the effectiveness. Finally, the performance evaluation of a variety of benchmark databases has been done by performance index. As a result, the proposed HBase approach outperforms the other databases.
For example, when sighted scholars study mathematics and physics etcetera, they need to access visual information, e.g., graphs and pictures. Furthermore, sighted people can express their own ideas and opinions visually. On the other hand, blind people can access visual information if it is expressed tactilely, but find it difficult to express their ideas and opinions visually. We are therefore developing a computer-aided system enabling blind people to draw their own figures on their own. This system consists of a matrix braille display to edit computer line drawings. The matrix braille display enables the blind to feel a tactile graphic during editing. After explaining two input methods for elementary plane shapes, we discuss two methods for scrolling tactile graphics to make the matrix braille display large enough to show tactile graphics in sufficient detail. We then show experimental results for using input and scrolling, and conclude with discussion on the usability of input and scrolling.
The trend toward more electric vehicles has demanded the need for high efficiency, high voltage and long life battery systems [1, 2]. Also renewable energy systems carry huge battery backups to overcome the renewable source shortage. Battery systems are affected by many factors, cells unbalancing is one of most important among these factors. Without the balancing system, individual cell voltages will differ over time that will decrease the battery pack capacity quickly. This condition is especially severe when the battery has a long string of cells and frequent regenerative charging is done via battery pack. Cell balancing is a method of designing safer battery solutions that extends battery runtime as well as battery life. Balancing mechanism can help in equalizing the state of charge across the multiple cells, therefore increasing the performance of battery system. Different cell balancing methodologies have been proposed for battery pack in recent years. These methods have some merits and demerits in comparison to each other; e.g. balancing time, complexity and active or passive balancing etc. In this paper, current bypass active cell balancing and Arduino based monitoring system designing and implementation is carried out. In charging process, this balancing technique provides partial current bypass using charging slope for weak cells. Also the passive shunt resistor technique is implemented to compare and verify the proposed system efficient response. Output result shows that this proposed balancing technique can perform cell balancing in much effective and efficient way as compared to previous balancing techniques. Using this cell balancing technique, we can improve overall battery health and lifetime.
Biped locomotion created by a controller based on Zero-Moment Point (ZMP) known as reliable control method looks different from human’s walking on the view point that ZMP-based walking does not include falling state, and it’s like monkey walking because of knee-bended walking profiles. However, the walking control that does not depend on ZMP is vulnerable to turnover. Therefore, keeping the event-driven walking of dynamical motion stable is important issue for realization of human-like natural walking. In this research, a walking model of humanoid robot including slipping, bumping, surface-contacting and line-contacting of foot is discussed, and its dynamical equation is derived by the Extended NE method. In this paper we introduce the humanoid model which including the slipping foot and verify the model.
In reinforcement learning, agents can learn appropriate actions for each situation based on the consequences of these actions after interacting with the environment. Reinforcement learning is compatible with self-organizing maps that accomplish unsupervised learning by reacting to impulses and strengthening neurons. Therefore, numerous studies have investigated the topic of reinforcement learning in which agents learn the state space using self-organizing maps. In this study, while we intended to apply these previous studies to transfer the learning and visualization of the human learning process, we introduced self-organizing maps into reinforcement learning and attempted to make their “state and action” learning process visible. We performed numerical experiments with the 2D goal-search problem; our model visualized the learning process of the agent.
The distribution of water pollution is often assessed by remote sensing. In this study, we develop a fuzzy multiple regression model and analyze water quality using data collected by the Advanced Visible and Near Infrared Radiometer type-2 (AVNIR-2) of the Advanced Land Observing Satellite at different time points. We conduct a fuzzy multiple regression analysis of the AVNIR-2 data and direct measurements of the local water quality of Lake Hachiroko in Japan. The relationship between the AVNIR-2 and water quality data are analyzed by solving both min and max problems. We compare the estimated water quality maps with the actual distributions in the study area, and determine that the proposed method enables us to derive water quality conditions effectively from the AVNIR-2 data. Furthermore, by comparing maps created using AVNIR-2 data collected at different times, we obtain results revealing temporal changes in water quality. In addition, we compare maps created using the fuzzy multiple regression and fuzzy regression models. We demonstrate that the former offers a greater number of solutions and provides more details about water quality.
With the publicized benefits offered by renewable energy resources, more and more households embrace the utilization of stand-alone installations ranging from small to medium scale systems. In literature, several studies provide insights on the effects of integration of renewable energy (RE) resources to the distribution systems but have inadequacy of considering the penetration levels. Moreover, RE cost reductions, increasing costs of traditional energy sources, and Renewable Portfolio Standards have created the possibility of significant increase of penetration levels of distributed RE generation being installed on distribution systems. To aid in the evaluation and assist with these expansions, new analysis tools are needed. In particular, new RE high-penetration analysis tools and procedures need to be developed and integrated with existing conventional methods. This paper presents a simulation based study on distribution system with and without integration of RE sources. It takes into account of the impending effects of these RE integrations in the distribution system. This paper emphasizes a novel method of determining the penetration level of Distributed Generation using least square minimization (LSM) method. The studies were tested using IEEE 123 bus distribution test feeder and actual data from an existing distribution system to verify the effectiveness and robustness of the proposed approach.
This paper proposes a novel approach to extracting local geometric features of the cultural relic. We first calculate Gaussian and mean curvature of the model. Then the surfaces of model are labeled for three fundamental types by the value of Gaussian and mean curvature. We use the region growing and expanding search method to obtain the local features of the model. We construct the templates based on the local features and prior knowledge. Finally, we achieve the retrieval of cultural relics by comparing the similarity of the template and the local feature of the model. We apply our method to identify and classify on Terracotta Warriors fragments. Experiments show that our method has the good retrieval performance.
The purpose of time-interval sequential pattern mining is to help superstore business managers promote product sales. Sequential pattern mining discovers the time interval patterns for items: for example, if most customers purchase product item A, and then buy items B and C after r to s and t to u days respectively, the time interval between r to s and t to u days can be provided to business managers to facilitate informed marketing decisions. We treat these time intervals as patterns to be mined, to predict the purchasing time intervals between A and B, as well as B and C. Nevertheless, little work considers the significance of product items while mining these time-interval sequential patterns. This work extends previous work and retains high-utility time interval patterns during pattern mining. This type of mining is meant to more closely reflect actual business practice. Experimental results show the differences between three mining approaches when jointly considering item utility and time intervals for purchased items. In addition to yielding more accurate patterns than the other two methods, the proposed UTMining_A method shortens execution times by delaying join processing and removing unnecessary records.