State-of The-art Approaches Research Articles

Machine Learning Based Prediction of Reference Evapotranspiration (ET0) Using IoT

Accurate estimation of Reference Evapotranspiration (ET₀) is important for efficient management and conservation of irrigation water. Existing methods of ET₀ rate determination are complex for application at the farmer level. Apart from standard methods of ET₀ determination, many data-driven soft computing approaches were also proposed to determine the ET₀ with limited data set. We proposed a temperature and humidity-based ML approach for ET₀ rate determination on directly sensed environmental conditions of the crop field. Crop field environmental conditions for (ET₀) rate determination are sensed by the proposed Internet of Things (IoT) architecture. Crop field environmental conditions from the year 2015 to 2021 in Pakistan are used for the training and testing of the proposed model. Gaussian Naive Bayes (GNB), Support Vector Machine (SVM), k-Nearest Neighbours (KNN), and Artificial Neural Network (ANN) based models are compared for performance. Crop fields directly sensed temperature and humidity pass to the model to train and predict the ET₀-rate of crop fields. The 10-fold cross-validation technique is applied for the evaluation of the proposed approach. The accuracy of the proposed solution for the ET₀ rate is compared against the Food and Agriculture Organization (FAO) recommended Penman-Monteith method for ET₀ rate determination. As concerned of the ML-based models the KNN model is more accurate as compared to SVM,GNB and ANN models with 92% accuracy. The KNN model of ET₀ is more efficient in reducing the Root Mean Squared Errors (RMSE) by 16% and Mean Absolute Errors (MAE) by 3% against the state of the art approach.

State of the Art Approach for Comprehensive Power System Security Assessment - Real Case Study

State of the Art Approach for Comprehensive Power System Security Assessment - Real Case Study

Towards Hybrid Energy-Efficient Power Management in Wireless Sensor Networks.

Wireless Sensor Networks (WSNs) are prone to highly constrained resources, as a result ensuring the proper functioning of the network is a requirement. Therefore, an effective WSN management system has to be integrated for the network efficiency. Our objective is to model, design, and propose a homogeneous WSN hybrid architecture. This work features a dedicated power utilization optimization strategy specifically for WSNs application. It is entitled Hybrid Energy-Efficient Power manager Scheduling (HEEPS). The pillars of this strategy are based on the one hand on time-out Dynamic Power Management (DPM) Intertask and on the other hand on Dynamic Voltage and Frequency Scaling (DVFS). All tasks are scheduled under Global Earliest Deadline First (GEDF) with new scheduling tests to overcome the Dhall effect. To minimize the energy consumption, the HEEPS predicts, defines and models the behavior adapted to each sensor node, as well as the associated energy management mechanism. HEEPS’s performance evaluation and analysis are performed using the STORM simulator. A comparison to the results obtained with the various state of the art approaches is presented. Results show that the power manager proposed effectively schedules tasks to use dynamically the available energy estimated gain up to 50%.

A Novel Single Anchor Localization Method for Wireless Sensors in 5G Satellite-Terrestrial Network

Satellites extend the 5G Internet of Things (IoT) applications to remote areas by providing global coverage. In applications like environmental and wildlife monitoring, the location information of IoT sensors is equally important as the data itself. This location information can also be utilized for spectrum management between the satellite and IoT networks. The localization is performed using nodes with known positions called anchor nodes (ANs). Optimization analysis of ANs significantly enhance the localization performance. In this paper, we propose a single AN-based localization algorithm for wireless sensors. The proposed localization approach uses directional information and mobility, focusing on finding the shortest path with only two beacon points. Simulation results demonstrate the superiority of the proposed localization algorithm over the state of the art approaches.

CCGraMi: An Effective Method for Mining Frequent Subgraphs in a Single Large Graph

In modern applications, large graphs are usually applied in the simulation and analysis of large complex systems such as social networks, computer networks, maps, traffic networks. Therefore, graph mining is also an interesting subject attracting many researchers. Among them, frequent subgraph mining in a single large graph is one of the most important branches of graph mining, it is defined as finding all subgraphs whose occurrences in a dataset are greater than or equal to a given frequency threshold. In which, the GraMi algorithm is considered the state of the art approach and many algorithms have been proposed to improve this algorithm. In 2020, the SoGraMi algorithm was proposed to optimize the GraMi algorithm and presented an outstanding performance in terms of runtime and storage space. In this paper, we propose a new algorithm to improve SoGraMi based on connected components, called CCGraMi (Connected Components GraMi). Our experiments on four real datasets (both directed and undirected) show that the proposed algorithm outperforms SoGraMi in terms of running time as well as memory requirements.

Automatic features extraction of faults in PEM fuel cells by a siamese artificial neural network

In this paper, a procedure aimed at the automatic extraction of the features from polymer electrolyte membrane fuel cell impedance spectra is proposed. An artificial neural network that is trained by exploiting the similarity learning concept has been used. The network learns the features of the impedance spectra and maps each of them into the embedding space by clustering them accurately and by emphasising differences among spectra corresponding to different faults. The siamese network structure is optimised and the quality of the learnt representation is evaluated by analysing the clusters obtained in the features space. The dataset of experimental spectra has been augmented in two different ways and the results are compared. The clustering quality of the proposed siamese network is compared with the one of other state of the art approaches.

Machine Learning Based Security Solutions in MANETs: State of the art approaches

Machine learning (ML) techniques provide the learning capability to a system and encourage adaptation into the environment, based upon many logical and statistical operations. The prime goal of ML is to recognize the complex patterns and make decisions based on the results. There are various ML algorithms which are implemented to secure the mobile ad-hoc networks. The infrastructure-less environment of MANETs poses a great challenge in implementation of the security systems. The security approaches in MANETs mainly focus on intrusion detection, malicious attacks mitigation, elimination of outlier/misbehavior/selfish nodes and securing routing paths. The researchers have been using cutting edge technologies for providing efficient security solutions by taking into the consideration of dynamic environment of MANETs. These technologies include machine learning, Artificial Intelligence (AI), Genetic Algorithms based methods, biological-inspired algorithms and so on. This paper presents a comprehensive and systematic study of various modern approaches for intensifying security in MANETs.

Demand Forecasting of Individual Probability Density Functions with Machine Learning

Demand forecasting is a central component of the replenishment process for retailers, as it provides crucial input for subsequent decision making like ordering processes. In contrast to point estimates, such as the conditional mean of the underlying probability distribution, or confidence intervals, forecasting complete probability density functions allows to investigate the impact on operational metrics, which are important to define the business strategy, over the full range of the expected demand. Whereas metrics evaluating point estimates are widely used, methods for assessing the accuracy of predicted distributions are rare, and this work proposes new techniques for both qualitative and quantitative evaluation methods. Using the supervised machine learning method "Cyclic Boosting", complete individual probability density functions can be predicted such that each prediction is fully explainable. This is of particular importance for practitioners, as it allows to avoid "black-box" models and understand the contributing factors for each individual prediction. Another crucial aspect in terms of both explainability and generalizability of demand forecasting methods is the limitation of the influence of temporal confounding, which is prevalent in most state of the art approaches.

Modeling of Visco-Elastic Environments for Humanoid Robot Motion Control

This manuscript presents a model of compliant contacts for time-critical humanoid robot motion control. The proposed model considers the environment as a continuum of spring-damper systems, which allows us to compute the equivalent contact force and torque that the environment exerts on the contact surface. We show that the proposed model extends the linear and rotational springs and dampers - classically used to characterize soft terrains - to the case of large contact surface orientations. The contact model is then used for the real-time whole-body control of humanoid robots walking on visco-elastic environments. The overall approach is validated by simulating walking motions of the iCub humanoid robot. Furthermore, the paper compares the proposed whole-body control strategy and state of the art approaches. In this respect, we investigate the terrain compliance that makes the classical approaches assuming rigid contacts fail. We finally analyze the robustness of the presented control design with respect to non-parametric uncertainty in the contact-model.

Pedestrian attribute recognition using trainable Gabor wavelets

Surveillance cameras are everywhere keeping an eye on pedestrians or people as they navigate through the scene. Within this context, our paper addresses the problem of pedestrian attribute recognition (PAR). This problem entails the extraction of different attributes such as age-group, clothing style, accessories, footwear style etc. This is a multi-label problem with a host of challenges even for human observers. As such, the topic has rightly attracted attention recently. In this work, we integrate trainable Gabor wavelet (TGW) layers inside a convolution neural network (CNN). Whereas other researchers have used fixed Gabor filters with the CNN, the proposed layers are learnable and adapt to the dataset for a better recognition. We test our method on publicly available challenging datasets and demonstrate considerable improvements over state of the art approaches.

Constraint Logic Programming for Real-World Test Laboratory Scheduling

The Test Laboratory Scheduling Problem (TLSP) and its subproblem TLSP-S are real-world industrial scheduling problems that are extensions of the Resource-Constrained Project Scheduling Problem (RCPSP). Besides several additional constraints, TLSP includes a grouping phase where the jobs to be scheduled have to be assembled from smaller tasks and derive their properties from this grouping. For TLSP-S such a grouping is already part of the input. In this work, we show how TLSP-S can be solved by Answer-set Programming extended with ideas from other constraint solving paradigms. We propose a novel and efficient encoding and apply an answer-set solver for constraint logic programs called clingcon. Additionally, we utilize our encoding in a Very Large Neighborhood Search framework and compare our methods with the state of the art approaches. Our approach provides new upper bounds and optimality proofs for several existing benchmark instances in the literature.

Fast Sidelobe Suppression Based on Two-Dimensional Joint Iterative Adaptive Filtering

In multitarget scenarios, the masking of small targets by large targets nearby may severely deteriorate radar detectability due to the range and Doppler sidelobes. In this article, a 2-D joint iterative adaptive filtering (2-D JIAF) method is proposed, by adopting the reiterative minimum mean square error (RMMSE) criterion to the outputs of a 2-D matched filter. The main advantages of the proposed method over the state-of-the-art (SOTA) methods, including modified adaptive multipulse compression and iterative adaptive approach, are twofold: 1) it is able to suppress the sidelobes in both range and Doppler dimensions and thus obtain an improved range-Doppler image; 2) the computational complexity is significantly reduced by adopting a small processing window in both range and Doppler dimensions. The derivation of 2-D JIAF is detailed and an efficient two-stage implementation is outlined. The performance of 2-D JIAF is validated with simulation results over a wide range of scenarios and compared with two SOTA approaches. The impacts of different parameters, including the number of iterations and the choice of the size of the processing window, are also extensively studied with Monte Carlo trials.

A Smart Hoodie- Assistive Technology for Blind People Using IoT

Assistive technology for the blind is a field of research that is gaining more and more importance due to the eruption of new interest from different fields. These research findings have had a huge social impact among elderly and blind population. In this paper, we introduced a novel Smart Hoodie which can sense the objects around them and can prevent the blind people from getting hurt accidentally or banging on heavy objects. The technology used here is IoT, where a Raspberry Pi kit is used along with ultrasonic sensors which can sense objects. A micro camera can be fixed on top of the Hoodie, which is triggered once the sensor detects the object and then object recognition is performed; a voice notification is generated either with the speaker or earphones. We can also track the location of a blind person in case they are lost, using a GPS module. A GSM module is used by a blind person to send messages in case of any emergency. Although, there were a lot of discoveries related to the above-mentioned issue, it has inflamed that our invention is much better than other state of the art approaches.

Decoding Imagined Speech and Computer Control using Brain Waves

BackgroundIn this work, we explore the possibility of decoding Imagined Speech (IS) brain waves using machine learning techniques. ApproachWe design two finite state machines to create an interface for controlling a computer system using an IS-based brain–computer interface. To decode IS signals, we propose a covariance matrix of Electroencephalogram channels as input features, covariance matrices projection to tangent space for obtaining vectors from matrices, principal component analysis for dimension reduction of vectors, an artificial neural network (ANN) as a classification model, and bootstrap aggregation for creating an ensemble of ANN models. ResultBased on these findings, we are first to use an IS-based system to operate a computer and obtain an information transfer rate of 21-bits-per-minute. The proposed approach can decode the IS signal with a mean classification accuracy of 85% on classifying one long vs. short word. Our proposed approach can also differentiate between IS and rest state brain signals with a mean classification accuracy of 94%. ComparisonAfter comparison, we show that our approach performs equivalent to the state-of-the-art approach (SOTA) on decoding long vs. short word classification task. We also show that the proposed method outperforms SOTA significantly on decoding three short words and vowels with an average margin of 11% and 9%, respectively. ConclusionThese results show that the proposed approach can decode a wide variety of IS signals and is practically applicable in a real-time environment.

Data augmentation and transfer learning to classify malware images in a deep learning context

In the past few years, malware classification techniques have shifted from shallow traditional machine learning models to deeper neural network architectures. The main benefit of some of these is the ability to work with raw data, guaranteed by their automatic feature extraction capabilities. This results in less technical expertise needed while building the models, thus less initial pre-processing resources. Nevertheless, such advantage comes with its drawbacks, since deep learning models require huge quantities of data in order to generate a model that generalizes well. The amount of data required to train a deep network without overfitting is often unobtainable for malware analysts. We take inspiration from image-based data augmentation techniques and apply a sequence of semantics-preserving syntactic code transformations (obfuscations) to a small dataset of programs to generate a larger dataset. We then design two learning models, a convolutional neural network and a bi-directional long short-term memory, and we train them on images extracted from compiled binaries of the newly generated dataset. Through transfer learning we then take the features learned from the obfuscated binaries and train the models against two state of the art malware datasets, each containing around 10 000 samples. Our models easily achieve up to 98.5% accuracy on the test set, which is on par or better than the present state of the art approaches, thus validating the approach.

A Queue-Stabilizing Framework for Networked Multi-Robot Exploration

Motivated by planetary exploration, we consider the problem of deploying a network of mobile robots to explore an unknown environment and share information with a stationary data sink. The configuration of robots affects both network connectivity and the accuracy of relative localization. Robots explore autonomously and can store data locally in their queues. When a communication path exists to the data sink, robots transfer their data. Because robots may fail in a non-deterministic manner, causing loss of the data in their queues, enabling communication is important. However, strict constraints on connectivity and relative positions limit exploration. To take a more flexible approach to managing these multiple objectives, we use Lyapunov-based stochastic optimization to maximize new information while using virtual queues to constrain time-average expectations of metrics of interest. These include queueing delay, network connectivity, and localization uncertainty. The result is a distributed online controller which autonomously and strategically breaks and restores connectivity as needed. We explicitly account for obstacle avoidance, limited sensing ranges, and noisy communication/ranging links with line-of-sight occlusions. We use queuing theory to analyze the average delay experienced by data in our system and guarantee connectivity will be recovered when feasible. We demonstrate in simulation that our queue-stabilizing controller can reduce localization uncertainty and achieve better coverage than two state of the art approaches.

Random area-perimeter method for generation of unimodal and multimodal cancelable biometric templates

In recent times, biometric based authentication systems have seen a tremendous growth in various applications. However, if databases in multiple applications are created using the same biometric characteristic and algorithm, then any compromise of the stored template in one biometric system may jeopardise the security of the other biometric systems as well. More importantly, such a compromise may also lead to a permanent loss of the biometric characteristic. Therefore, the cancelability or revocability of biometrics has become quite an essential requirement. In this paper, a novel scheme, the Random Area & Perimeter Method (RAPM)) is presented in which a biometric characteristic of an individual is transformed into random values which are stored as cancelable biometric templates. The proposed scheme computes area and perimeter of the Bezier curve which are obtained through interpolation of feature points of original biometrics and a random point chosen by the user. The area and perimeter thus computed exhibit pseudo-random properties. This technique outperforms many existing techniques. Moreover, a dimensionality reduction to the tune of more than 95% has been obtained without compromising the matching performance. The average values obtained for EER, DI and RI are 0.0045, 6.28 and 99.64 respectively. These values are better than those obtained from the available state of the art approaches. The proposed scheme has also been analysed for various privacy issues like non-invertibility, revocability, and unlinkability.

A Graph-Theoretic Approach for Spatial Filtering and Its Impact on Mixed-Type Spatial Pattern Recognition in Wafer Bin Maps

Statistical quality control in semiconductor manufacturing hinges on effective diagnostics of wafer bin maps, wherein a key challenge is to detect how defective chips tend to spatially cluster on a wafer--a problem known as spatial pattern recognition. Recently, there has been a growing interest in mixed-type spatial pattern recognition--when multiple defect patterns, of different shapes, co-exist on the same wafer. Mixed-type spatial pattern recognition entails two central tasks: (1) spatial filtering, to distinguish systematic patterns from random noises; and (2) spatial clustering, to group filtered patterns into distinct defect types. Observing that spatial filtering is instrumental to high-quality mixed-type pattern recognition, we propose to use a graph-theoretic method, called adjacency-clustering, which leverages spatial dependence among adjacent defective chips to effectively filter the raw wafer maps. Tested on real-world data and compared against a state-of the-art approach, our proposed method achieves at least 46% gain in terms of internal cluster validation quality (i.e., validation without external class labels), and about ~5% gain in terms of Normalized Mutual Information--an external cluster validation metric based on external class labels. Interestingly, the margin of improvement appears to be a function of the pattern complexity, with larger gains achieved for more complex-shaped patterns.

Data-Driven Frequency Dynamic Unit Commitment for Island Systems With High RES Penetration

The replacement of directly connected synchronous generators with power electronics interfaced generation has led to a decrease in system's inertia posing a significant challenge on frequency dynamics. In isolated systems with reduced inertia predefined limits for renewable penetration and primary reserves are frequently set for dynamic security purposes. This approach might not ensure dynamic security or can prove conservative in certain conditions. Furthermore, these approaches rarely consider the capabilities of inverter based renewable generation to provide frequency services. In this paper, a data driven approach, based on optimal classification trees is proposed to extract, from a detailed dynamic model of the system, the constraints for a frequency dynamic unit commitment formulation. Hence, both dynamic security and optimal exploitation of renewable and conventional units for power production and frequency support can be achieved. The advantages of the proposed method compared to conventional and state of the art approaches in frequency security are validated through dynamic simulations on a realistic model of Rhodes island and IEEE 118. Uncertainties in load demand and renewable generation are dealt by a robust optimization method. Its economic performance, computational overhead and modelling complexity is compared to a stochastic approach.

A novel sequential synthesis algorithm for the integrated optimization of Rankine cycles and heat exchanger networks

The research theme of this work is the optimization of the heat integration and heat recovery in energy systems and chemical processes, specifically, the optimal synthesis and design of Rankine cycles (e.g., heat recovery cycles, heat pump cycles, etc.) integrated with the heat exchanger network. The challenging synthesis problem is formulated as a Mixed Integer NonLinear Program and tackled with a novel sequential algorithm based on the idea of optimizing the independent mass flow rates of the Rankine cycle superstructure with a derivative-free algorithm. At the lower level, for fixed Rankine cycle and utility mass flow rates, the synthesis of the heat exchanger network is performed with an efficient sequential algorithm. The proposed algorithm is compared against three state-of the-art approaches on six real-world case studies including Organic Rankine cycles, heat pumps/CHP cycles and heat recovery steam cycles with multiple pressure levels. Results indicate that the proposed algorithm finds optimal or near-optimal design solutions for all case studies proving its applicability as an effective design tool.