Xiaoying Kong

INS algorithm using quaternion model for low cost IMU

Abstract This paper presents a generic inertial navigation system (INS) error propagation model that does not rely on small misalignment angles assumption. The modelling uses quaternions in the computer frame approach. Based on this model, an INS algorithm is developed for low cost inertial measurement unit (IMU) to solve the initial attitudes uncertainty using in-motion alignment. The distribution approximation filter (DAF) is used to implement the non-linear data fusion algorithm.

Wireless sensor network localization algorithm using dynamic path of mobile beacon

Localization is one of the basic prerequisites of sensors in various applications of wireless sensor networks. A beacon is a special sensor with geographical knowledge and which can be employed to help localize general sensors. A mobile beacon is treated as a replacement for many static beacons since it is movable and flexible and often powerful. The path of a mobile beacon will determine the rate of coverage and accuracy of position determination that it supports. Whereas a static path is planned before any localization action, a dynamic path is determined in real-time based on the demands of initially unknown sensors and hence can be more efficient than a static path. In this paper, we proposed a method of localization employed a mobile beacon whose path will be selected according to the real-time information of unknown sensors. The method is designed to be both “thrifty” in both energy consumption and economical cost, and lightweight in terms of computation load. Simulation results show that the method is lightweight but effective and efficient.

BSPNN: boosted subspace probabilistic neural network for email security

In the modern age of Internet connectivity, advanced information systems have accumulated huge volumes of data. Such fast growing, tremendous amount of data, collected and stored in large databases has far exceeded our human ability to comprehend without proper tools. There has been a great deal of research conducted to explore the potential applications of Machine Learning technologies in Security Informatics. This article studies the Network Security Detection problems in which predictive models are constructed to detect network security breaches such as spamming. Due to overwhelming volume of data, complexity and dynamics of computer networks and evolving cyber threats, current security systems suffer limited performance with low detection accuracy and high number of false alarms. To address such performance issues, a novel Machine Learning algorithm, namely Boosted Subspace Probabilistic Neural Network (BSPNN), has been proposed which combines a Radial Basis Function Neural Network with an innovative diversity-based ensemble learning framework. Extensive empirical analyses suggested that BSPNN achieved high detection accuracy with relatively small computational complexity compared with other conventional detection methods.

NavOptim Coding: Supporting Website Navigation Optimisation using Effort Minimisation

Web applications have rapidly become critical to the interaction that organisations have with their external stakeholders. A major factor in the effectiveness of this interaction is the ease with which navigation within the application can occur, and especially the extent to which users can locate information and functionality which they are seeking. Effective design is however complicated by the multiple purposes and users which Web applications typically support. Despite the fact that this implies that navigation design is inherently an optimisation problem, few optimisation techniques have been applied in this domain - with most design techniques being based on intuition, general heuristics, or experimental refinement. In this paper we discuss this problem, and propose a navigation representation which can become the basis for optimisation techniques.

Dynamic Path Determination of Mobile Beacons Employing Reinforcement Learning for Wireless Sensor Localization

Wireless sensor networks (WSN) are extensively applied in civil and military areas. Localization is an essential prerequisite for many WSN applications, and is often based on beacons that provide geographical information in real time. Mobile Beacons (MB) can be used to replace many static beacons with paths that can be controlled in real-time. Robotic and/or flight vehicles can work as MBs. In this paper we consider the use of reinforcement learning (RL) (a significant branch of machine learning) to control MBs. Usually, RL needs an infinite series of episodes to determine an optimal policy. We propose however a method of localization employing mobile beacon whose behavior will be controlled by an adapted RL algorithm. A MB learns and makes decisions based on weighted information collected from unknown sensors. Simulation results show that the adapted RL algorithm provides sufficient information to the MB to localise unknown sensors in a lightweight but effective way.

GPS Modeling in Frequency Domain

Although DGPS provides positioning information with high precision, when DGPS is unavailable in some situations, stand-alone GPS has to be used in vehicle navigation. The accuracy of standard GPS is low due to position measurement errors. This paper presents a frequency domain modeling approach to model GPS errors and increase GPS positioning accuracy. This approach models GPS errors using shaping filter. External sensors are employed to reduce GPS errors. This paper also presents an approach to select external sensors to meet the quality requirements of positioning system.

How project duration, upfront costs and uncertainty interact and impact on software development productivity? A simulation approach

Identifying impact factors on software development productivity and the static relations between the impact factors and performance has been the main focus in the literature. Insight into the dynamic relation between key factors and performance dimensions would expand and complement the conventional wisdom on software development productivity. This is the first study to present such dynamic relationship based on an Analytical Theory of Project Investment. Through simulation, we have demonstrated the dynamic relationship between project duration, the uncertainty level of the perceived project value, the fixed project upfront cost and software development productivity. The findings provide practitioners with insight into how these factors interact and impact on software development project productivity.

A shaping filter approach to model GPS errors

Although Differential Global Positioning System (DGPS) provides positioning information with high precision, when DGPS is unavailable in some situations, stand-alone GPS has to be used in vehicle navigation. The accuracy of standard GPS is low due to position measurement errors. This paper presents a frequency domain modelling approach to model GPS errors and increase GPS positioning accuracy. This approach models GPS errors using shaping filter. External sensors are employed to reduce GPS errors. This paper also presents an approach to select external sensors to meet the accuracy requirements of positioning system.

Wireless Sensor Network Localization with Autonomous Mobile Beacon by Path Finding

Sensor data with geographical information become ubiquitous with development of location base applications in various businesses. There is no exception to Wireless sensor network (WSN) as it is getting more practical under the booming of semiconductor industry. Localization of WSN usually depends on beacons that equip with GPS or GSM module. The cost of finance and energy hamper the application of WSN. Robot or flight vehicle called Mobile Beacon (MB) can relieve the issue and enhance the application area of WSN. Based on the framework we proposed in [14], we renovate the algorithm to find the best position for MB in every step by grouping weight of cosine similarity to mining the relation between responsible sensors. The MB machine works in an autonomous mode to find dynamic path without any supervision. In simulation, it is proved that the novel method makes the best of observation and covers unknown sensors as many as possible without heavy computation and struggle of energy. With appropriate parameters, the method can reach a 90% average coverage rate on any random distribution of sensors.

Performance comparison of packet scheduling algorithms in LTE-A HetNets

Performance comparison of various Packet Scheduling (PS) algorithms such as Proportional Fair (PF), Maximum Largest Weighted Delay First (MLWDF) and Exponential/Proportional Fair (EXP/PF) has been studied in HetNets environment. The performance indicators such as throughput, Packet Loss Ratio (PLR), delay and fairness are considered to judge the performance of the scheduling algorithms. Various strategies such as increasing the number of pico cells in the cell edge were used in the simulation for the performance evaluation study. The results achieved by various simulations show that adding the pico cells to the existing macros enhances the overall system performance in addition to various scheduling algorithms implemented in macros. Simulation results show that the overall system gain has increased by adding picos, provide better coverage in the cell edge and increase the capacity of the network to provide better Quality of Service (QoS). Furthermore, simulations show that MLWDF performs better for video traffic than compared to other with PS algorithms.