Co-cluster extraction is a basic approach for summarization of cooccurrence information. This paper proposes a visual assessment technique for co-cluster structure analysis through cooccurrence-sensitive ordering, which realizes the hybrid concept of the coVAT algorithm and distance-sensitive ordering in relational data clustering. Object-item cooccurrence information is first enlarged into an (object + item) × (object + item) cooccurrence data matrix, and then, cooccurrence-sensitive ordering is performed through spectral ordering of the enlarged matrix. Additionally, this paper also consider the intuitive validation of co-cluster structures considering cluster crossing curves, which was adopted in cluster validation with distance-sensitive ordering. The characteristic features of the proposed approach are demonstrated through several numerical experiments including application to social analysis of Japanese prefectural statistics.
Multi-sensor fusion and target tracking are two key technologies for the environmental awareness system of autonomous vehicles. In this paper, a moving target tracking method based on the fusion of Lidar and binocular camera is proposed. Firstly, the position information obtained by the two types of sensors is fused at decision level by using adaptive weighting algorithm, and then the Joint Probability Data Association (JPDA) algorithm is correlated with the result of fusion to achieve multi-target tracking. Tested at a curve in the campus and compared with the Extended Kalman Filter (EKF) algorithm, the experimental results show that this algorithm can effectively overcome the limitation of a single sensor and track more accurately.
Detection and tracking of dynamic obstacle is one of the research hotspot in autonomous vehicles. In this paper, a dynamic obstacle detection and tracking method based on 3D lidar is proposed. The nearest neighborhood method is used to cluster the data obtained by the laser lidar. The characteristic parameters of the clustering obstacles are analyzed. Multiple hypothesis tracking model (MHT) algorithm and the nearest neighbor association algorithm are used for data association of two consecutive frames of obstacle information. The dynamic and static state of obstacles are analyzed through the temporal and spatial correlation of the obstacle. Finally, we use linear Kalman filter to predict the movement state of the obstacle. The experimental results on a low-speed driverless vehicle “small whirlwind” which is an autonomous sightseeing vehicle show that the method can accurately detect the dynamic obstacles in unknown environment with effectiveness and real-time performance.
Cross-media retrieval has raised a lot of research interests, and a significant number of works focus on mapping the heterogeneous data into a common subspace using a couple of projection matrices corresponding to each modal data before implementing similarity comparison. Differently, we reconstruct one modal data (e.g., images) to the other one (e.g., texts) using a model named sparse neural network pre-trained by Restricted Boltzmann Machines (MRCR-RSNN) so that we can project one modal data into the space of the other one directly. In the model, input is low-level features of one modal data and output is the other one. And cross-media retrieval is implemented based on the similarities of their representatives. Our model need not any manual annotation and its application is more widely. It is simple but effective. We evaluate the performance of our method on several benchmark datasets, and experimental results prove its effectiveness based on the Mean Average Precision (MAP) and Precision Recall (PR).
Task allocation is an important concept not only in biological systems but also in artificial systems. This paper reports a case study of autonomous task allocation behavior in an evolutionary robotic swarm. We address a path-formation task that is a fundamental task in the field of swarm robotics. This task aims to generate the collective path that connects two different locations by using many simple robots. Each robot has a limited sensing ability with distance sensors, a ground sensor, and a coarse-grained omnidirectional camera to perceive its local environment and the limited actuators composed of two colored LEDs and two-wheeled motors. Our objective is to develop a robotic swarm with autonomous specialization behavior from scratch, by exclusively implementing a homogeneous evolving artificial neural network controller for the robots to discuss the importance of embodiment that is the source of congestion. Computer simulations demonstrate the adaptive collective behavior that emerged in a robotic swarm with various swarm sizes and confirm the feasibility of autonomous task allocation for managing congestion in larger swarm sizes.
The stability regions of a LCL-filtered converter adopting converter-current-feedback control without damping are analyzed. The nonlinear LCL-filtered model is presented to investigate its influence on the system stability. The stability analysis is performed by means of the Nyquist diagram in s domain. It reveals that three factors have the dominant effects on the system stability, including internal loss of LCL-filtered model, PWM transport delay and controller parameters. The undamped stability boundaries of the system gain calculated by the symmetrical optimum method are obtained. It can be found that stable regions for the nonlinear LCL-filtered system are extended into a continuous region of ratios of LCL filter resonance frequency to control frequency from three distinct regions. Finally, the stable regions are validated by the nonlinear model simulation, and experimental results verify the theoretical analysis.
Organizations can be considered as complex systems that can adapt to their changing environment. In this work, we study a complex system adapting to an unfamiliar environment with learning; this is grounded in the context of the post-acquisition integration of the companies. More specifically, we conceptualize post-acquisition integration from the perspective of behavioral theory as a reason for the environmental changes to the firms (agents). We studied the adaptation of these complex systems and we propose a coupled learning method over the NK landscape. The simulation results show that the initial perceptions of the agents regarding the new task environment can be quite influential to the performance of the entire system during the adaptation process. Correct initial perceptions can help the system to quickly achieve high performance, whereas incorrect initial perceptions may prevent the system from reaching high performance. Lack of initial perceptions could lead to a slow yet robust adaptation process with a moderate level of performance. Moreover, certain other factors, such as the sensitivity to the feedback from the environment, the incentive of the system for exploration, and the learning frequency, may have different impact on the adaptation and performance of the system.
The focus of this paper is to explore the use of kernel combinations of the support vector machines (SVMs) for vehicle classification. Being the primary component of the SVM, the kernel functions are responsible for the pattern analysis of the vehicle dataset and to bridge its linear and non-linear features. However, the choice of the type of kernel functions has characteristics and limitations that are highly dependent on the parameters. Thus, in order to overcome these limitations, a method of compounding kernel function for vehicle classification is hereby introduced and discussed. The vehicle classification accuracy of the compound kernel function presented is then compared to the accuracies of the conventional classifications obtained from the four commonly used individual kernel functions (linear, quadratic, cubic, and Gaussian functions). This study provides the following contributions: (1) The classification method is able to determine the rank in terms of accuracies of the four individual kernel functions; (2) The method is able to combine the top three individual kernel functions; and (3) The best combination of the compound kernel functions can be determined.
Detection and tracking of road lane markings offers several applications in intelligent transport systems (ITS). Although it is perceived as the simple task of isolating lanes on various types of roads, the accuracy of detection remains an issue. Several studies in recent literature have proposed solutions to this problem; however, none of these have used the method of macroblock (MB) prediction. This paper focuses on the type of MB applied for lane detection, tracking, and predictions, as well as the trade-off between the accuracy and complexity of implementing the system. This study makes the following contributions: (1) best MB for spatiotemporal lane detection and reconstruction; (2) best function approximation for lane predictions; and (3) best MB in terms of performance under different conditions.
The Tsallis entropy is a q-parameter extension of the Shannon entropy. By maximizing it within the framework of fuzzy c-means, statistical mechanical membership functions can be derived. We propose a clustering algorithm that includes the membership function and deterministic annealing. One of the major issues for this method is the determination of an appropriate values for q and an initial annealing temperature for a given data distribution. Accordingly, in our previous study, we investigated the relationship between q and the annealing temperature. We quantitatively compared the area of the membership function for various values of q and for various temperatures. The results showed that the effect of q on the area was nearly the inverse of that of the temperature. In this paper, we analytically investigate this relationship by directly integrating the membership function, and the inversely proportional relationship between q and the temperature is approximately confirmed. Based on this relationship, a q-incrementation deterministic annealing fuzzy c-means (FCM) algorithm is developed. Experiments are performed, and it is confirmed that the algorithm works properly. However, it is also confirmed that differences in the shape of the membership function of the annealing method and that of the q-incrementation method are remained.
The design of a comfortable and functional prosthetic hand is still a challenge. This paper presents the design of a tendon-driven, 3D-printed, underactuated prosthetic hand. An improved structural design was developed to make the hand more flexible. Three fingers are equipped with abduction freedom at the metacarpophalangeal joints (MCP) to ensure natural enveloping for both cylinder and sphere-like objects. A force-sensing resistor (FSR) is adopted to measure the fingertip force of each finger. Experiments show that this type of structure design provides the hand with excellent dexterity, as the added abduction ensures natural enveloping grasp gestures for both cylinder and sphere-like objects. Moreover, a myoelectric control paradigm is implemented in the control system to demonstrate the feasibility.
We herein investigate the influence of object detection in deep learning. Based on using one neural network model and maintaining its primary network structure, we discuss the relationship between the detection accuracy with the scale of the training dataset and the network depth and width. We adopt the single factor experiment for each influence factor and create a test dataset including different types of object pictures. After each experiment, we first predict the average precision for the validation dataset and subsequently test the target pictures. The results of the experiment reveal that it is effective to improve the accuracy by enriching the training dataset. The more necessary features the training dataset has, the more precise are the results. Therefore, the network structure is a crucial factor, and adopting advanced models could be beneficial to obtain an excellent performance on sophisticated targets.
Image similarity measures play an important role in nearest neighbor search and duplicate detection for large-scale image datasets. Recently, Minwise Hashing (or Minhash) and its related hashing algorithms have achieved great performances in large-scale image retrieval systems. However, there are a large number of comparisons for image pairs in these applications, which may spend a lot of computation time and affect the performance. In order to quickly obtain the pairwise images that theirs similarities are higher than the specific threshold T (e.g., 0.5), we propose a dynamic threshold filter of Minwise Hashing for image similarity measures. It greatly reduces the calculation time by terminating the unnecessary comparisons in advance. We also find that the filter can be extended to other hashing algorithms, on when the estimator satisfies the binomial distribution, such as b-Bit Minwise Hashing, One Permutation Hashing, etc. In this pager, we use the Bag-of-Visual-Words (BoVW) model based on the Scale Invariant Feature Transform (SIFT) to represent the image features. We have proved that the filter is correct and effective through the experiment on real image datasets.
This paper presents fuzzy logic algorithm for Power Efficiency, Reliable Irrigation, and Temperature Control (PERITC) of Smart Farm. The desired motor speed for pump irrigation and required temperature inside the plant chamber is obtained using fuzzy logic system. The fuzzy logic inputs are data from the water reservoir, plant water requirements, power optimization control, inside temperature, and outside temperature of the plant chamber. The results of this study show that the controller using fuzzy algorithm are reliable, efficient and robust.
This paper presents the development of a vision-based system for microscopic road traffic scene analysis and understanding using computer vision and computational intelligence techniques. The traffic flow model is calibrated using the information obtained from the road-side cameras. It aims to demonstrate an understanding of different levels of traffic scene analysis from simple detection, tracking, and classification of traffic agents to a higher level of vehicular and pedestrian dynamics, traffic congestion build-up, and multi-agent interactions. The study used a video dataset suitable for analysis of a T-intersection. Vehicle detection and tracking have 88.84% accuracy and 88.20% precision. The system can classify private cars, public utility vehicles, buses, and motorcycles. Vehicular flow of every detected vehicles from origin to destination are also monitored for traffic volume estimation, and volume distribution analysis. Lastly, a microscopic traffic model for a T-intersection was developed to simulate a traffic response based on actual road scenarios.
Human activity recognition with the smartphone could be important for many applications, especially since most of the people use this device in their daily life. A smartphone is a portable gadget with internal sensors and enough hardware power to accommodate this problem. In this paper, three neural network algorithms were compared to detect six major activities. The data are collected by a smartphone in real life and simulated on the remote server. The results show that MLP and GMDH neural network have better accuracy and performance compared with the LVQ neural network algorithm.
A fuzzy logic-based controller with fuzzy rule formulation algorithm on software and actual control for a 6-DOF robotic arm was implemented. A robotic arm with 4-DOF attached a 2-DOF gripper serves as the testing platform. The actual robotic arm was characterized and the parameters are used for the simulator to mimic actual response. The fuzzy logic controller is then implemented to the simulated robotic arm and was then implemented to the actual robotic arm to control its movement. The new features are as follows: (1) Implementation of simulated weight and frictional effects of dynamic robot arm movement, (2) formulation and justification of reduced fuzzy rules for control and (3) addition of a path-partitioning algorithm to further enhance the dynamic movement of the robotic arm.
One of the challenges in image quality assessment (IQA) is to determine the quality score without the presence of the reference image. In this paper, the authors proposed a no-reference image quality assessment method based on the natural statistics of double-opponent (DO) cells. It utilizes the statistical modeling of the three opponency channels using the generalized Gaussian distribution (GGD) and asymmetric generalized Gaussian distribution (AGGD). The parameters of GGD and AGGD are then applied to feedforward neural network to predict the image quality. Result shows that for any opposing channels, its natural statistics parameters when applied to feedforward neural network can achieve satisfactory prediction of image quality.
Recently, local mascot characters called “Yurukyara” have been active in various places. They play an important role in raising a region’s image and exciting the region to promote regional development. It is important to understand the impression given by the characters, since each character’s image leads to the promotion and recognition of the region. In this study, we analyze the impression of local mascot characters to provide useful information for regional promotions, etc. First, we extract the Kansei factors from the characters’ appearances and classify the characters within the factor space. Next, we analyze the differences in impressions when adding character-profile and video information.
On September 6, 2017, we lost a great researcher, Prof. Lotfi A. Zadeh, the one who introduced one of the most important mathematical concepts that gets a good rapport with reality. “Fuzzy Sets” was the paper he published in 1965. Many researchers and practitioners of mathematics, science, engineering, medicine, and economics found it fascinating. In Japan, consumer electronics incorporating ambiguity in human thought and behavior became popular. As these products became big topics, the term “fuzzy” also became popular in the 1990s.
This set of papers offers a sample of the expanding development of fuzzy logic and soft computing. One review paper is written by Prof. Takeshi Yamakawa, an internationally famous fuzzy researcher of fuzzy hardware systems. Four of these papers were selected through a peer review process. One, “Design of Fuzzy Logic Controller and its Distinctive Features” by Prof. Takeshi Yamakawa, describes the design of a fuzzy logic controller and its application to controlling mouse-platform stabilization. The distinctive features of fuzzy logic control are also discussed.
The first original paper, “MMMs-Induced Possibilistic Fuzzy Co-Clustering and its Characteristics” by Ubukata et al., describes an NPCCMM scheme by considering variable cluster volumes and the fuzziness degree of item memberships to investigate the specific aspects of fuzzy rather than probabilistic nature in co-clustering tasks. The second original paper, “Inner Evaluation of Writing in a Foreign Language Based on Expert Judgment for Correction” by Entani et al., proposes a method of numerically deriving each writer’s writing preference from an expert’s initial evaluation. Third, “A 3-Dimensional Fuzzy Linguistic Evaluation Model” by Suprasongsin et al., describes a new model for determining criteria weights in group decision-making problems, a model based on the concept of probabilistic linguistic terms. The fourth original paper, “Natural Language Generation Using Monte Carlo Tree Search” by Kumagai et al., proposes a method for generating simulation-based natural language. The method accounts for both building a correct syntactic structure and reflecting the given situational information as input in the sentences generated.
We appreciate Prof. Takeshi Yamakawa for the excellent review of fuzzy controllers. We also thank all of the authors who submitted original articles and all of the reviewers who offered their valuable comments and suggestions.
Prof. Lotfi A. Zadeh, who created a new approach to describe a knowledge of a human expert with a natural language, passed away on September 6, 2017. His significant accomplishment was to create a novel artificial intelligence (AI) which exhibits the knowledge of human experts in natural linguistic terms. This system is structured and clear in two points of why a result is obtained and how it is done. The system contrasts with AI systems based on neural networks or deep learning.
In this paper, the design of a fuzzy logic controller and its application to controlling of the mouse-platform stabilization are described. In addition, the distinctive features of fuzzy logic control are discussed. The author wants to offer this paper on the altar of Prof. Zadeh.
In the field of cluster analysis, fuzzy theory including the concept of fuzzy sets has been actively utilized to realize flexible and robust clustering methods. Fuzzy C-means (FCM), which is the most representative fuzzy clustering method, has been extended to achieve more robust clustering. For example, noise FCM (NFCM) performs noise rejection by introducing a noise cluster that absorbs noise objects and possibilistic C-means (PCM) performs the independent extraction of possibilistic clusters by introducing cluster-wise noise clusters. Similarly, in the field of co-clustering, fuzzy co-clustering induced by multinomial mixture models (FCCMM) was proposed and extended to noise FCCMM (NFCCMM) in an analogous fashion to the NFCM. Ubukata et al. have proposed noise clustering-based possibilistic co-clustering induced by multinomial mixture models (NPCCMM) in an analogous fashion to the PCM. In this study, we develop an NPCCMM scheme considering variable cluster volumes and the fuzziness degree of item memberships to investigate the specific aspects of fuzzy nature rather than probabilistic nature in co-clustering tasks. We investigated the characteristics of the proposed NPCCMM by applying it to an artificial data set and conducted document clustering experiments using real-life data sets. As a result, we found that the proposed method can derive more flexible possibilistic partitions than the probabilistic model by adjusting the fuzziness degrees of object and item memberships. The document clustering experiments also indicated the effectiveness of tuning the fuzziness degree of object and item memberships, and the optimization of cluster volumes to improve classification performance.
Although writing is a tool for communication, the way one writer communicates a fact is not always the same as how another one does it. The written word is unique to the writer and reflects his or her preferred writing style. When something is written by a non-native speaker of language, native speakers and experts often feel slightly unusual, even if they can find no obvious errors. Moreover, they might revise the text based on their experience. On the other hand, the writer often feels slightly dissatisfied with the correction if it does not fit for his or her writing preference. It is difficult for the corrector to understand the writers’ writing preference from the text, and it is also difficult for the writer to explain it explicitly since both writing and correcting a piece of text are based on one’s subjectivity. The correction is unique to the text, so the inner evaluation of the text is important. This study proposes a method of deriving each writer’s writing preference numerically from the expert’s initial evaluation. In the process, the texts other than the target text are taken into consideration from the viewpoint that writing is a communication tool. The corrector may use the feedback from the proposed method to confirm his or her intuitive judgments and to add some new viewpoints.
A probabilistic linguistic-based model is an effective tool to express preferences with different weights for different linguistic terms. This paper aims at introducing a new model for determining criteria weights in group decision-making problems, which is based on the concept of probabilistic linguistic terms. Different linguistic weights of respondents are also incorporated into the proposed model. Fuzzy numbers are used to quantify the linguistic terms. Using this model, first, a new concept called three-dimensional fuzzy linguistic representation is proposed to serve as an extension of the existing models. Then, a normalization process, an aggregation process, and a defuzzifying process for three-dimensional fuzzy linguistic representation are investigated. Next, a model for determining criteria weights is formulated. A case study of a beverage product in Thailand is provided to demonstrate the applicability of the proposed model. Finally, the results are compared with the existing models.
We propose a method of simulation-based natural language generation that accounts for both building a correct syntactic structure and reflecting the given situational information as input for the generated sentence. We employ the Monte Carlo tree search for this nontrivial search problem in simulation, using context-free grammar rules as search operators. We evaluated numerous generation results from two aspects: the appropriateness of sentence contents for the given input information and the sequence of words in a generated sentence. Furthermore, in order to realize an efficient search in simulation, we introduced procedures to unfold syntactic structures from words strongly related to the given situational information, and increased the probability of selecting those related words. Through a numbers of experiments, we confirmed that our method can effectively generate a sentence with various words and phrasings.