Jose Santos

A Survey of Geographical Routing in Wireless Ad-Hoc Networks

Geographic routing offers a radical departure from previous topology-dependent routing paradigms through its use of physical location in the routing process. Geographic routing protocols eliminate dependence on topology storage and the associated costs, which also makes them more suitable to handling dynamic behavior frequently found in wireless ad-hoc networks. Geographic routing protocols have been designed for a variety of applications ranging from mobility prediction and management through to anonymous routing and from energy efficiency to QoS. Geographic routing is also part of the larger area of context-awareness due to its usage of location data to make routing decisions and thus represents an important step in the journey towards ubiquitous computing. The focus of this paper, within the area of geographic routing is on wireless ad-hoc networks and how location information can benefit routing. This paper aims to provide both a comprehensive and methodical survey of existing literature in the area of geographic routing from its inception as well as acting as an introduction to the subject.

An evaluation of indoor location determination technologies

The development of real-time locating systems (RTLS) has become an important add-on to many existing location aware systems. While GPS has solved most of the outdoor RTLS problems, it fails to repeat this success indoors. A number of technologies have been used to address the indoor tracking problem. The ability to accurately track the location of people indoors has many applications ranging from medical, military and logistical to entertainment. However, current systems cannot provide continuous real-time tracking of a moving target or lose capability when coverage is poor. The deployment of a real-time location determination system however is fraught with problems. To date there has been little research into comparing commercial systems on the market with regard to informing IT departments as to their performance in various aspects which are important to tracking devices and people in relatively confined areas. This article attempts to provide such a useful comparison by providing a review of the practicalities of installing certain location-sensing systems. We also comment on the accuracies achieved and problems encountered using the position-sensing systems.

SWAT: A Spiking Neural Network Training Algorithm for Classification Problems

This paper presents a synaptic weight association training (SWAT) algorithm for spiking neural networks (SNNs). SWAT merges the Bienenstock-Cooper-Munro (BCM) learning rule with spike timing dependent plasticity (STDP). The STDP/BCM rule yields a unimodal weight distribution where the height of the plasticity window associated with STDP is modulated causing stability after a period of training. The SNN uses a single training neuron in the training phase where data associated with all classes is passed to this neuron. The rule then maps weights to the classifying output neurons to reflect similarities in the data across the classes. The SNN also includes both excitatory and inhibitory facilitating synapses which create a frequency routing capability allowing the information presented to the network to be routed to different hidden layer neurons. A variable neuron threshold level simulates the refractory period. SWAT is initially benchmarked against the nonlinearly separable Iris and Wisconsin Breast Cancer datasets. Results presented show that the proposed training algorithm exhibits a convergence accuracy of 95.5% and 96.2% for the Iris and Wisconsin training sets, respectively, and 95.3% and 96.7% for the testing sets, noise experiments show that SWAT has a good generalization capability. SWAT is also benchmarked using an isolated digit automatic speech recognition (ASR) system where a subset of the TI46 speech corpus is used. Results show that with SWAT as the classifier, the ASR system provides an accuracy of 98.875% for training and 95.25% for testing.

Clinical use of automatic pacemaker algorithms: results of the AUTOMATICITY registry.

AIMS Follow-up of the ever-increasing numbers of patients with implantable cardiac devices places a heavy burden on clinical departments. Device automaticity may alleviate the follow-up burden by minimizing the time for physician involvement. The aim of the prospective, multicentre AUTOMATICITY registry was to examine the performance of a subset of programmed automatic algorithms during patient follow-up and their acceptance by implanting physicians. METHODS AND RESULTS The clinical use of automatic algorithms from the Insignia pacemakers (PM; Boston Scientific, St Paul, MN, USA) was evaluated: atrial and ventricular AutoSense (sensitivity adjustment), ventricular Automatic Capture (threshold verification and output setting), AutoLifeStyle (sensor settings adjustment). The objective of the study was to assess the reprogramming rates within 12 months of implant, the reasons for reprogramming and relationship to adverse events. A total of 960 patients were enrolled in the study. The proportion of patients free from any algorithm reprogramming at 12 months was 86.1%. A total of 2736 algorithms were activated at enrolment, with 156 (5.7%) being reprogrammed in 115 patients at 12 months for any reason. Forty-nine reprogrammings (1.8%) were unintentional or related to changes in device settings such that the algorithm was no longer available, 33 (1.2%) were due to suspected sensing issues, and 22 (0.8%) were assumed related to the algorithm. The individual 12-month reprogramming-free rates were: ventricular AutoSense 94.3%, Atrial AutoSense 93.3%, AutoLifeStyle 93.9%, and Automatic Capture 95.9%. CONCLUSION The results of the AUTOMATICITY registry show that automatic measurement of key settings and automatic adjustment to optimal programming is feasible and safe. The simplicity of PM follow-up and avoidance of frequent reprogramming may contribute to a more effective use of hospital time and resources.

MANET location prediction using machine learning algorithms

In mobile ad-hoc networks where users are potentially highly mobile, knowledge of future location and movement can be of great value to routing protocols. To date, most work regarding location prediction has been focused on infrastructure networks and consists of performing classification on a discrete range of cells or access points. Such techniques are unsuitable for infrastructure-free MANETs and although classification algorithms can be used for specific, known areas they are not general or flexible enough for all real-world environments. Unlike previous work, this paper focuses on regression-based machine learning algorithms that are able to predict coordinates as continuous variables. Three popular machine learning techniques have been implemented in MATLAB and tested using data obtained from a variety of mobile simulations in the ns-2 simulator. This paper presents the results of these experiments with the aim of guiding and encouraging development of location-predictive MANET applications.

Towards a location and mobility-aware routing protocol for improving multimedia streaming performance in MANETs

The increasing availability and decreasing cost of mobile devices equipped with WiFi radios has led to increasing demand for multimedia applications in both professional and entertainment contexts. The streaming of multimedia however requires strict adherence to QoS levels in order to guarantee suitable quality for end users. MANETs lack the centralised control, coordination and infrastructure of wireless networks as well as presenting a further element of complexity in the form of device mobility. Such constraints make achieving suitable QoS a nontrivial challenge and much work has already been presented in this area. This paper proposes a bottom-up routing protocol which specifically takes into account mobility and other unique characteristics of MANETs in order to improve QoS for multimedia streaming. Geographic Predictive Routing (GPR) uses Artificial Neural Networks to accurately predict the future locations of neighbouring devices for making location and mobility-aware routing decisions. GPR is intended as the first step towards creating a fully QoS-aware networking framework for enhancing the performance of multimedia streaming in MANETs. Simulation results comparing GPR against well-established ad-hoc routing protocols such as AODV and DSR show that GPR is able to achieve an improved level of QoS in a variety of multimedia and mobility scenarios.

SWAT: An unsupervised SNN training algorithm for classification problems

The work presented in this paper merges the Bienenstock-Cooper-Munro (BCM) learning rule with spike timing dependent plasticity (STDP) to develop a training algorithm for a spiking neural network (SNN), stimulated using spike trains. The BCM rule is utilised to modulate the height of the plasticity window, associated with STDP. The SNN topology uses a single training neuron in the training phase where all classes are passed to this neuron, and the associated weights are subsequently mapped to the classifying output neurons: the weights are proportionally distributed across the output neurons to reflect similarities in the input data. The training algorithm also includes both exhibitory and inhibitory facilitating dynamic synapses that create a frequency routing capability allowing the information presented to the network to be routed to different hidden layer neurons. A variable neuron threshold level simulates the refractory period. The network is benchmarked against the non-linearly separable IRIS data set problem and results presented in the paper show that the proposed training algorithm exhibits a convergence accuracy comparable to other SNN training algorithms.

A time multiplexing architecture for inter-neuron communications

This paper presents a hardware implementation of a Time Multiplexing Architecture (TMA) that can interconnect arrays of neurons in an Artificial Neural Network (ANN) using a single metal wire. The approach exploits the relative slow operational speed of the biological system by using fast digital hardware to sequentially sample neurons in a layer and transmit the associated spikes to neurons in other layers. The motivation for this work is to develop minimal area inter-neuron communication hardware. An estimate of the density of on-chip neurons afforded by this approach is presented. The paper verifies the operation of the TMA and investigates pulse transmission errors as a function of the sampling rate. Simulations using the Xilinx System Generator (XSG) package demonstrate that the effect of these errors on the performance of an SNN, pre-trained to solve the XOR problem, is negligible if the sampling frequency is sufficiently high.

CAPTURE - Extending the scope of self-localization in Indoor Positioning Systems

The concept of devices cooperatively assisting with the localization of other devices in either the indoor or outdoor arena is not a new phenomenon. The primary focus of research into such a theory is however, limited to solving the problem of localization accuracy. In this work, the motivation is to provide a solution to the current range limitations of an Indoor Position System (IPS) in the form of a framework “Cooperatively Applied Positioning Techniques Utilizing Range Extension (CAPTURE)”. These range limitations are the culmination of well documented difficulties of localizing using wireless signals in such Non-Line of Sight (NLOS) environments. The coverage of a localization solution is still a new and challenging issue in the indoor environment. In this paper we implement a version of CAPTURE that uses Wi-Fi Direct and Bluetooth Low Energy (Bluetooth LE 4.0) that take advantage of mobile devices at the outer limits of an IPS to help extend its reach into blind spots, where devices cannot be currently located. CAPTURE is evaluated using a live test environment, where range estimations are recorded between cooperating devices. These range estimations are filtered before being placed into a positioning algorithm to locate lost devices. Finally the accuracy of CAPTURE is presented, demonstrating the achievable benefits of implementing CAPTURE as a solution to the problem of coverage in an Indoor environment.

An Empirical Evaluation of Robust Gaussian Process Models for System Identification

System identification comprises a number of linear and nonlinear tools for black-box modeling of dynamical systems, with applications in several areas of engineering, control, biology and economy. However, the usual Gaussian noise assumption is not always satisfied, specially if data is corrupted by impulsive noise or outliers. Bearing this in mind, the present paper aims at evaluating how Gaussian Process (GP) models perform in system identification tasks in the presence of outliers. More specifically, we compare the performances of two existing robust GP-based regression models in experiments involving five benchmarking datasets with controlled outlier inclusion. The results indicate that, although still sensitive in some degree to the presence of outliers, the robust models are indeed able to achieve lower prediction errors in corrupted scenarios when compared to conventional GP-based approach.