Tao Wang

Bio: Tao Wang is an academic researcher from Beijing Information Science & Technology University. The author has contributed to research in topics: Cluster analysis & Extended Kalman filter. The author has an hindex of 6, co-authored 21 publications receiving 83 citations. Previous affiliations of Tao Wang include Beihang University.

A Novel Measurement Data Classification Algorithm Based on SVM for Tracking Closely Spaced Targets

Abstract: Target tracking is an important field of investigation in wireless sensor networks. When multiple targets are closely spaced, their measurement points are mixed together due to the insufficient accuracy of the sensor. This may bring some difficulties in determining the positions of observation points required by subsequent algorithms and applications. Most of the known tracking algorithms are derived from the Kalman filter, extended Kalman filter, and particles filter. In this paper, a novel measurement data classification algorithm based on support vector machine (SVM) is provided. SVM and Kalman filter are combined to obtain the updated classification line at each sampling period, and the sampling points would be classified by the updated classification line to calculate the coordinates of the corresponding observation points, which are then used to estimate the precise positions of two targets. A series of simulations and experiments are carried out to validate the presented algorithm on classifying and tracking two targets moving closely. Simulation results for a maneuvering targets classification example illustrate the feasibility of the new algorithm, as well as experimental and quantitative results from the practical data validate the effectiveness and stability of our proposal in contrast with existing methods.

Measurement Data Classification Optimization Based on a Novel Evolutionary Kernel Clustering Algorithm for Multi-Target Tracking

Abstract: In the field of range-based multi-target localization and tracking, measurement data of each time cannot associate with its corresponding target, and clustering analysis can be used to solve this problem. In this paper, a novel evolutionary kernel clustering algorithm was developed for range-based multi-target tracking in wireless sensor networks. First, the locations of multi-targets are predicted according to the previous trajectories. Second, we apply the clustering number recognition algorithm to filter out the outliers and calculate the initial cluster center by analyzing the density of each measurement data. For each cluster, its relationship with corresponding target is established according to the predictive position and the initial cluster center. Third, the density factors of each measurement data are fused into the Gaussian kernel function to improve the accuracy of the cluster center. Finally, the accurate position of each target at current moment is calculated based on the predictive position and the measurement data set of corresponding cluster. Two different experiments are done in this paper: in the first experiment, the clustering performance of our proposed algorithm is evaluated based on the training data set. In the second one, the accuracy improvement of range-based target tracking is shown when proposed algorithm is used to analyze the measurement data. The tracking results show that the proposed algorithm is more robust to large localization errors.

Measurement Data Fusion Based on Optimized Weighted Least-Squares Algorithm for Multi-Target Tracking

Abstract: The outliers remove, the classification of effective measurements, and the weighted optimization method of the corresponding measurement are the main factors that affect the positioning accuracy based on range-based multi-target tracking in wireless sensor networks. In this paper, we develop an improved weighted least-square algorithm based on an enhanced non-naive Bayesian classifier (ENNBC) method. According to the ENNBC method, the outliers in the measurement data are removed effectively, dataset density peaks are found quickly, and remaining effective measurements are accurately classified. The ENNBC method improves the traditional direct classification method and took the dependence among continuous density attributes into account. Four common indexes of classifiers are used to evaluate the performance of the nine methods, i.e., the normal naive Bayesian, flexible naive Bayesian (FNB), the homologous model of FNB (FNB ROT ), support vector machine, k-means, fuzzy c-means (FCM), possibilistic c-means, possibilistic FCM, and our proposed ENNBC. The evaluation results show that ENNBC has the best performance based on the four indexes. Meanwhile, the multi-target tracking experimental results show that the proposed algorithm can reduce the root-mean-squared error of the position compared with the extended Kalman filter. In addition, the proposed algorithm has better robustness against large localization and tracking errors.

Lane Determination of Vehicles Based on a Novel Clustering Algorithm for Intelligent Traffic Monitoring

Abstract: In intelligent traffic monitoring, speed measuring millimeter waves (MMW) radar is one of the most commonly used tools for traffic enforcement. In traffic enforcement field, the radar must provide the evidence of each vehicle belongs to which lane. In this paper, we propose a novel kernel line segment adaptive possibilistic c-means clustering algorithm (KLSAPCM) for lane determination of vehicles. Firstly, the raw measurement data is preprocessed using the extracting method of data adjacent lane centerlines. Secondly, according to the improved minimum radius data search method, outliers are removed and the proposed KLSAPCM algorithm is initialized. Finally, the accuracy of lane determination has been improved by the proposed KLSAPCM clustering algorithm based on adaptive kernel line segment that conforms to the shape features of the measurement data in the actual scene. The experiment results for multiple scenes were: the KLSAPCM algorithm is compared with the DBSCAN, the $k$ -means, the FCM, the PCM, the AMPCM, and the APCM algorithms on real measurement datasets, and the results highlight the classification rate of the proposed algorithm. Meanwhile, the proposed algorithm gets a good real-time performance and strong robustness for some sparse moving vehicle scene applications.

Machine learning

Abstract: Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Energy-Aware Geographic Routing for Real-Time Workforce Monitoring in Industrial Informatics

Abstract: Workforce monitoring is a vital activity in large factories in order to oversee the worker’s concentration on their duty and increase productivity. Workforces are kind of moving targets which can be monitored via wireless sensor networks (WSNs). As sensor nodes have a limited source of energy, optimal energy consumption is of crucial importance in these networks. Several protocols for routing are designed in order to consider efficient energy consumption in conjunction with target tracking and coverage. In this article, a new energy-efficient routing algorithm geographic routing time transfer (GRTT) is proposed to use topological information of sensor nodes for target tracking and coverage applications. In this article, a weight called relay ability is defined for each node according to the sensor network topology. These weights are calculated and announced to sensor nodes by cluster heads (CHs). Once a target enters the area covered by sensor nodes, a signal is sent to the CH through the route having maximum predefined weights in the network. Simulations show better results than other tracking routing methods based on the metrics of energy consumption of the network, power consumption, and throughput for GRTT (proposed method), dynamic energy-efficient routing protocol (DEER), virtual force-based energy-hole mitigation (VFEM), nonequal-probability multicast routing protocol (MRP-NEP), and trace-announcing routing scheme (TARS) methods.

Machine Learning Techniques and A Case Study for Intelligent Wireless Networks

Abstract: With the widespread deployment of wireless technologies and IoT, 5G wireless networks will support various communication connectivity and services for the huge number of wireless smart/ intelligent devices and machines. The challenge lies in assisting wireless networks to intelligently learn experience, autonomously optimize network configurations and smartly make decisions to support massive wireless smart devices with minimum human intervention, so the diverse and colorful service requirements can be satisfied with the optimum performance. Machine learning, as one of the powerful artificial intelligence tools, is capable of efficiently supporting wireless smart devices by assisting them to smartly observe the environment, analyze data and make decisions with the intelligence. Hence, in this article, we briefly review the major concepts of common machine learning techniques and present their potential applications in intelligent wireless networks, including spectrum sensing, channel estimation, device clustering, behavior prediction, position tracking, data demission reduction, adaptive routing, energy harvesting/efficiency, resource management, and so on. Furthermore, we propose deep reinforcement learning for intelligent resource management in intelligent wireless networks in an exemplary case study. Simulation results demonstrate the effectiveness and advance of machine learning in intelligent wireless networks.