Intelligent Systems and Applications

As a follow-up to an earlier investigation, a true forward test has been carried out by applying a previously developed financial predictor (in the form of a so called pattern set, optimized using an evolutionary algorithm) to a new data set, involving data for 200 stocks and covering a time period from February 2016 to the end of that year. Despite being applied to previously unseen data, the pattern set generated a set of trades with an average one-day return of 0.394%. Moreover, the pattern set’s total trading return (excluding transaction costs) over the entire period covered by the new data, when applied as a trading strategy with a simple m–day holding period for each trade, was 15.9% for \(m=1\), 24.9% for \(m=3\), and 61.6% for \(m=6\), compared to 16.2% for the benchmark index (S&P 500) over the same period.

The present work introduces a new approach to the optimum wells placement problem in oil fields using evolutionary computation. In particular, our contribution is twofold: we propose an efficient algorithm for initialisation of highly constrained optimisation problems based on Monte-Carlo sampling and we propose a new optimisation technique that uses this population sampling scheme, a space-efficient chromosome and the application of cellular genetic algorithms to promote a large population diversity. Usually, authors define a domain representation having oil wells placed at any arbitrary position of the chromosome. On the other hand, the proposed representation enforces a unique relative wells position for each combination of wells. Therefore, the suggested scheme diminishes the problem size, thus making the optimisation more efficient. Moreover, by also employing a cellular genetic algorithm, we guarantee an improved population diversity along the algorithm execution. The experiments with the UNISIM-I reservoir indicate an enhancement of 6 to 10 times of the final NPV when comparing the proposed representation and the traditional one. Besides, the cellular genetic algorithm with the suggested chromosome performs better than the classical genetic algorithm by a factor of 1.5. The proposed models are valuable not only for the oil and gas industry but also to every integer optimisation problem that employs evolutionary algorithms.

Voltage Source Converters (VSCs) based Flexible AC Transmission Systems (FACTS) are popular for speedy regulation of different network parameters, thus being a strong candidate for transient stability enhancement by damping Low Frequency Oscillations (LFOs). Static Synchronous Series Compensator (SSSC) is a series FACTS controller with built in capability to absorb or deliver reactive power. SSSC may damp LFOs by installation of efficient supplementary damping control (SDC). Due to recent advancements in the field of Soft Computing (SC), there is a growing realization of their contribution to damping control design for FACTS. The direct focus of this chapter is to exploit the potential of a hybrid control, obtained from assorted domains such as NeuroFuzzy and Sliding Mode Control (SMC). SMC technique is the most lucrative choice to design SDC due to its optimal performance, delivery in critical applications with low complexity and high precision. The contributions of this framework are the damping performance improvement for single and multimachine power system with fast convergence speed.

The problem of robotic path planning is relevant to many applications that have led to extensive study. This has only increased as autonomous robotic vehicles have become more affordable and varied. Optimal solutions to this problem can be computationally expensive, leading to the need for efficiently achievable approximate solutions. In this work, we present a genetic algorithm to solve the path planning problem. The algorithm operates offline but runs onboard the micro aerial vehicle (MAV). This is accomplished by mounting a single-board computer on the vehicle and integrating it with the flight control board. In addition, we evaluate the effectiveness of two genetic operators: crossover and mass extinction. Results demonstrate that a standard, single-point crossover operator is largely ineffective. Mass extinction, an operator that has been used rarely in previous work, is explored within the framework of a genetic algorithm utilizing only mutation and selection. Based on initial results, mass extinction may have some utility for path planning, however, due to large number of parameters and potential implementations, additional experimentation is needed.

Methods from the area of Process Mining traditionally focus on extracting insight in business processes from event logs. In this paper we explore the potential of Process Mining to provide valuable insights in (un)healthy habits and to contribute to ambient assisted living solutions when applied on data from smart home environments. Events in smart home environments are recorded at the level of sensor triggers, which is too low to mine habit-related behavioral patterns. Process discovery algorithms produce then overgeneralizing process models that allow for too much behavior and that are difficult to interpret for human experts. We show that abstracting the events to a higher-level interpretation can enable discovery of more precise and more comprehensible models. We present a framework to automatically abstract sensor-level events to their interpretation at the human activity level. Our framework is based on the XES IEEE standard for event logs. We use supervised learning techniques to train it on training data for which both the sensor and human activity events are known. We demonstrate our abstraction framework on three real-life smart home event logs and show that the process models that can be discovered after abstraction improve on precision as well as on F-score.

In modern production lines, smaller batches to be produced and higher customization level of a single component bring to higher cost, related especially to setup and preparation of machines. The setup of a milling machine is an operation that requires time and may bring to errors that can be catastrophic. In this Chapter, the ARTool Augmented Reality framework for machine tool operations is presented. The framework permits to write and debug part-code in an augmented environment, to identify quicker misalignments and errors in fixing of new blank material, and to support maintenance operations. The ego-localization of the handheld device that depicts the augmented scene in machine work-area is based upon markers. The library that performs marker identification is brand-new and it is benchmarked throughout the Chapter against a state-of-the-art solution (ARUCO) and a ground truth (multi-stereoscopic motion capture). The Chapter also describes the general information flow and the context that brought to the conception of the ARTool framework, and presents a series of applications developed using the framework.

New dynamical systems with strange attractors are numerically investigated in the article. These dynamical systems correspond to the main mathematical model describing the attitude dynamics of multi-spin spacecraft and gyrostat-satellites. The considering dynamical systems are structurally related to the well-known Newton-Leipnik system. Properties of the strange attractors arising inside the phase spaces of the dynamical systems are examined with the help of the numerical modelling.

It can be seen that a huge attention has been given to the field of location-aware systems. Different technologies such as: RFID, NFC, QR, GPS and iBeacon, have been developed to be used for these systems. However, in order to design an efficient system for location-based content delivery, there is a need to study these technologies with their features and issues. Moreover, this study provides a review and comparison for the available mobile applications that use iBeacon technology for delivering contextual content. According to the result of evaluation and comparison, the idea is developing a location-aware mobile system using the iBeacon technology. This system can help the event’s organizers to enhance attendee’s experience by providing location based content such as PDFs, image and videos. This is can provide an efficient way for content delivery by reducing the costs and replacing the printed papers.

Artificial Social Companions are a promising solution for the increasing challenges in elderly care. This chapter describes the CaMeLi autonomous conversational agent system which simulates human-like affective behaviour and acts as a companion for older adults living alone at home. The agent employs synthetic speech, gaze, facial expressions, and gestures to support multimodal natural interaction with its users and assists them in a number of daily life scenarios.We present the agent’s overall architecture, with a focus on the perception, decision making and synthesis components which give rise to the agent’s intelligent affective behavior. The agent was evaluated in an exploratory study where it was introduced in 20 homes of older adults (aged 65+) in three European countries (Switzerland, the Netherlands, Portugal) for a total duration of 12 weeks. We present the results of the evaluation study with regards to acceptance, perceived usability, and usefulness of the agent, and discuss future opportunities for fellow researchers who are striving to bring virtual agents out of the laboratories into successful real world applications.

This research proposed working with an influence on the subject mood, presenting an approach to state the subjects analysis when the light hue is varied. The experimental results led to the finding of the emotional response time dynamics. Such dynamics are important for future design and implementation of the control loops in-house automation systems for emotion modulation. Throughout this document, the details and progress of the research in emotional domotics, with the aim of developing a controlled algorithm for living space based on the user’s emotional state, will be illustrated and detailed. This project is centered on domotics (home automation) systems, which is, a set of elements installed, interconnected and controlled by a computer system. After introducing the investigation’s core, general preview, and the experiment’s description conducted with light hue variation, the description is followed by a presentative approach to state the subjects analysis when light hue is varied. The experimental results led to the time dynamics of emotional response findings. Such dynamics are important for future design and implementation of the control loops in house automation systems for emotion modulation.

In the public goods game, players can be classified into different types according to their participation in the game. It is an important issue for economists to be able to measure players’ strategy changes over time which can be considered as concept drift. In this study, we present a method for measuring changes in items’ cluster membership in temporal data. The method consists of three steps in the first step, the temporal data will be transformed into a discrete series of time points then each time point will be clustered separately. In the last step, the items’ membership in the clusters is compared with a reference of behaviour to determine the amount of behavioural change in each time point. Different external cluster validity indices and area under the curve are used to measure these changes. Instead of different cluster label comparison, we use these indices a new way to compare between clusters and reference points. In this study, three categories of reference of behaviours are used 1- first time point, 2- previous time pint and 3- the general overall behaviour of the items. For the public goods game, our results indicate that the players are changing over time but the change is smooth and relatively constant between any two time points.

The neuromorphic visual processing inspired by the biological vision system of brain offers an alternative process into applying machine vision in various environments. With the emerging interests on transportation safety enhancement of Advanced Driver Assistance System or a driverless car, the neuromorphic convolutional recurrent neural networks was proposed and tested for the night-time vehicle or VRU detection. The effectiveness of proposed convolutional-recurrent neural networks of neuromorphic visual processing was evaluated successfully for the object detection without optimized complex template matching or prior denoising neural network. The real life road video dataset at night time demonstrated 98% of successful detection/segmentation rate with 0% False Positive. The robust performance of proposed convolutional-recurrent neural network was also applied successfully to the tooth segmentation of dental X-ray 3D CT including the gum region. The feature extraction was based on neuromorphic visual processing filters of either hand-cut filters mimicking the visual cortex experimentation or the auto-encoder filter trained by partial X-ray images. The consistent performance of either hand-cut filters or the small auto-encoder filters demonstrated the feasibility of real-time and robust neuromorphic vision implemented by either the small embedded system or the portable computer.

Game theory and its application in multi-agent systems continues to attract a considerable number of scientists and researchers around the globe. Moreover, the need for distributed resource allocation is increasing at a high pace and multi-agent systems are known to be suitable to deal with these problems. In this chapter, we investigate the presence of multiple resources in minority games where each resource can be given a weight (importance). In this context, we investigate different settings of the parameters and how they change the results of the game. In spite of some previous works on multi-resource minority games, we explain why they should be referred as multi-option games. Through exploring various scenarios of multi-resource situations, we take into account two important issues: (i) degree of freedom to choose strategy, and (ii) the effect of resource capacity on the different evaluation criteria. Besides, we introduce a new criterion named resource usage to understand the behavior of the system and the performance of agents in utilizing each resource. We find that although using a single strategy may involve less computation, using different strategies is more effective when employing multiple resources simultaneously. In addition, we investigate the system behavior as the importance of resources are different; we find that by adjusting the weight of resources, it is possible to attract agents towards a particular resource.

In recent years, the number of machine-to-machine (M2M) networks that do not require direct human intervention has been increasing at a rapid pace. However, the need for a wireless platform to control and monitor these M2M networks, one with both a vast coverage area and a low network deployment cost, continues to be unmet. Wireless Sensor Networks (WSNs) with energy efficiency routing protocols in M2M environments are emerging to meet the challenges of such communication through network convergence. M2M communication is considered as the core of the Internet of Things (IoT). IoT refers to a network of billions of objects that can send and receive data. Energy efficiency, delay are a critical issue in M2M and there is a shortfall in IP addresses in IoT. In this chapter, an energy efficient routing protocol for Wireless Sensor Networks (WSN) is presented, which provides a platform to control and M2M networks. Inefficient energy consumption caused by nodes being active all the time is tackled using an adaptive sleep mode solution to maintain high levels of Network Performance (N.P). Firstly, a Multilevel Clustering Multiple Sink (MLCMS) with IPv6 protocol over Low Wireless Personal Area Networks (6LoWPAN) is promoted using a sophisticated mathematical equation for electing cluster heads (CH) for each level, so as to prolong network lifetime. Secondly, enhanced N.P that prolongs the life time of the system and maximises the reduction of delay is achieved through an adaptive sleep mode scheme. The sensor field is divided into quarters with different levels of cluster heads (CHs) and two optimal location sinks. The performance of the MLCMS protocol is evaluated and compared with the multi-hop low-energy adaptive clustering hierarchy (M-LEACH) protocol. MLCMS performs 62% better than M-LEACH and 147% more effectively regarding energy efficiency. Next, 6LoWPAN for the proposed model is constructed, and its impact on the performance of MLCMS by Network Simulator (NS3) simulation is evaluated. This increases the packets received by the system by 7% more than using MLCMS without 6LoWPAN and it improves the flexibility of the proposed model. Subsequently, an adaptive sleep mode scheme, based on CH’s residual energy for the active period time, is introduced for MLCMS and a comparative analysis establishes that it extends the lifetime of the system twice as much as the evaluated MLCMS without the adaptive sleep mode algorithm. Furthermore, with the sleep mode algorithm, this reduces the delay by a half and increases the delivery by 10%.

This paper proposes and analyses the applicability of integrating Fuzzy C-Means (FCM) and artificial neural network (ANN) in rainfall forecasting. The algorithm of ANN and FCM clustering are integrated and applied to forecast short-term localized rainfall in tropical weather. Rainfall forecasting in this paper is divided into state forecast (raining or not raining) and rainfall rate forecast. Various type of back propagation extended network with hidden layers of ANN structured were trained. Training algorithm of Levenberg-Marquardt, Bayesian Regularization and Scaled Conjugate Gradient are used and trained. Transfer function in each neuron uses linear, logistic sigmoid and hyperbolic tangent sigmoid. Initial statistical analysis of weather parameter, data pre-processing approach and FCM clustering method were used to organize input data for the ANN forecast model. Input parameters such as atmospheric pressure, temperature, dew point, humidity and wind speed have been used. One to six hour predicted rainfall forecast are compared and analyzed. The result indicates that the integrated of FCM-ANN forecast model yield 80% for 1 h forecast.

This work proposes the analysis between two neural network configurations for development a intelligent recognition system of speech signal patterns of numerical commands in Brazilian Portuguese. Thus, the Multilayer Perceptron (MLP) and Learning Vector Quantization (LVQ) networks are evaluated their performance in the course of training, validation and testing in speech signal recognition, whose pattern of speech signal is given by a two-dimensional time matrix, resulting of the encoding of the mel-cepstral coefficients (MFCC) through application of discrete cosine transform (DCT). These patterns have reduced set of parameters and the configurations of neural network in analysis use few examples for each pattern through training. It was carried out many simulations for network topologies and some selected learning algorithms to determine the network structures with best hit and generalization results. The potential this proposed approach is shown by check up on obtained outcomes with others classifiers, represented by Gaussian Mixture Models (GMM) and Support Vector Machines (SVM).

This paper firstly reviews five artificial intelligence tools that might be useful in helping tele-operators to drive mobile robots: knowledge-based systems (including rule based systems and case-based reasoning), automatic knowledge acquisition, fuzzy logic, neural networks and genetic algorithms. Rule-based systems were selected to provide real time support to tele-operators with their steering because the systems allow tele-operators to be included in the driving as much as possible and to reach their target destination, while helping when needed to avoid an obstacle. A bearing to an end-point is added as an input with an obstacle avoidance sensor system and the usual inputs from a joystick. A recommended direction is combined with the angle and position of a joystick and the rule-based scheme generates a recommended angle to rotate the mobile robot. That recommended angle is then blended with the user input to assist tele-operators with steering their robots in the direction of their destinations.

The development of a GPS based position control system for Wave Adaptive Modular Vessel (WAM-V) able to navigate between waypoints is discussed in this chapter. A fuzzy reasoned double loop controller is proposed for navigation path planning of WAM-V. For outer loop fuzzy controller is used to feed the desired heading to the inner loop. In the inner loop, a PID feedback controller is used to correct the desired course generated by the fuzzy reasoned algorithm. The control system provides the required feedback signals to track the desired heading which is obtained from the fuzzy algorithm. After PID generates the appropriate command, the thrust isallocated to the port side and starboard side thrusters along with the command from lookup table. Using the proposed controller, several experiments are conducted at Osaka University free running pond facility. The WAM-V is equipped without rudder, thus it is driven by a combination of different thrusts to control both speed and heading. Several experimental results with different sets of waypoint validate the proposed algorithm. The obtained results affirmed that the proposed fuzzy waypoint guidance control algorithm is powerful to realize the navigation path planning. The waypoint navigation experimental results show that the fuzzy guided waypoint controller scheme is simple, intelligent and robust. The goal of this research is to present a solution to the waypoint control problem for the underactuated catamaran vessel (WAM-V), which is achieved successfully.

Attaining high security in mobile ad hoc networks (or MANETs) is the utmost concern in the present era of wireless ad-hoc communication and efforts are continuously being made by the researchers, in order to provide a feasible solution for the same. Impervious security has to be ensured in MANETs because of their involvement in transacting highly sensitive information. However, it has been discerned that during the communication process, the decentralized and dynamic nature of MANETs impedes the security of mobile nodes. This study is an attempt to enhance the security in routing techniques of MANETs by overhauling the existing security system after its critical evaluation. It has been observed that the cryptographic techniques in use prove to be inefficient or fail in some of the current scenarios. Therefore, a non-cryptographic method has been put forward that strengthens the process of authenticating nodes in MANETs by taking into account two factors viz. trust and resource, unlike the conventional ones. On analyzing the performance with respect to throughput, packet delivery ratio, end-to-end delay and computational time, the proposed system proves to be better than the previous standard secure routing scheme.

The paper proposes a novel chaos based audio steganography and cryptography method. It is a higher Least Significant Bit (LSB) layers algorithm in which the secret message is encrypted first by one-time pad algorithm. Two chaotic sequences of Piecewise Linear Chaotic Map (PWLCM) were used. In the encryption process, the key for one-time pad is generated by PWLCM chaotic map. In the steganography process, the second sequence of PWLCM is used to generate a random sequence. Then, indices of the ordered generated sequence were used to embed the encrypted message in randomly selected audio samples. The encrypted data were embedded on the higher layers other than the LSB using efficient bits adjustment algorithm, in order to increase the robustness against noise addition or MPEG compression. An analysis is discussed for the proposed scheme. For the steganography algorithm, the proposed scheme overcomes the main two problems for LSB coding, which are the low robustness of secret message extraction and destruction. For the former, the proposed method encrypts the secret message by perfect efficient algorithm which is the one-time pad. Regard of the second one, the secret message is hidden in higher layers which improve the robustness against signal processing manipulation. The main three steganography characteristics were tested and evaluated which are high capacity, perceptual transparency and robustness. Furthermore, the drawbacks of key generation and key distribution for one-time pad is sloved by using the chaotic maps. For the experimental results, waveform analysis and signal-to-noise Ratio are made, which show the high quality of the stego audio, and hence demonstrate the efficiency of the proposed scheme.

This chapter presents a generalized and versatile connected health solution for patient monitoring. It consists of a mobile system that can be used at home, an ambulance and a hospital. The system uses the Shimmer sensor device to collect three axes (x, y and z) accelerometer data as well as electrocardiogram signals. The accelerometer data is used to implement a fall detection system using the k-Nearest Neighbors classifier. The classification algorithm is implemented on various platforms including a PC and the Zynq system on chip platform where both programmable logic and processing system of the Zynq are explored. In addition, the electrocardiogram signals are used to extract vital information, the signals are also encrypted using the Advanced Encryption Standard and sent wirelessly using Wi-Fi for further processing. Implementation results have shown that the best overall accuracy reaches 90% for the fall detection while meeting real-time performances when implemented on the Zynq and while using only 48% of Look-up Tables and 22% of Flip-Flops available on chip.