Konstantinos P. Ferentinos

Adaptive design optimization of wireless sensor networks using genetic algorithms

We present a multi-objective optimization methodology for self-organizing, adaptive wireless sensor network design and energy management, taking into consideration application-specific requirements, communication constraints and energy-conservation characteristics. A precision agriculture application of sensor networks is used as an example. We use genetic algorithms as the optimization tool of the developed system and an appropriate fitness function is developed to incorporate many aspects of network performance. The design characteristics optimized by the genetic algorithm system include the status of sensor nodes (whether they are active or inactive), network clustering with the choice of appropriate clusterheads and finally the choice between two signal ranges for the simple sensor nodes. We show that optimal sensor network designs constructed by the genetic algorithm system satisfy all application-specific requirements, fulfill the existent connectivity constraints and incorporate energy-conservation characteristics. Energy management is optimized to guarantee maximum life span of the network without lack of the network characteristics that are required by the specific application.

A drift-tolerant model for data management in ocean sensor networks

Traditional means of observing the ocean, like fixed mooring stations and radar systems, are difficult and expensive to deploy and provide coarse-grained and data measurements of currents and waves. In this paper, we explore the use of inexpensive wireless drifters as an alternative flexible infrastructure for fine-grained ocean monitoring. Surface drifters are designed specifically to move passively with the flow of water on the ocean surface and they are able to acquire sensor readings and GPS-generated positions at regular intervals. We view the fleet of drifters as a wireless ad-hoc sensor network with two types of nodes:i) a few powerful drifters with satellite connectivity, acting as mobile base-stations, and ii)a large number of low-power drifters with short-range acoustic or radio connectivity. Using real datasets from the Gulf of Maine (US) and the Liverpool Bay (UK), we study connectivity and uniformity properties of the ad-hoc mobile sensor network. We investigate the effect of deployment strategy, weather conditions as well as seasonal changes on the ability of drifters to relay readings to the end-users,and to provide sufficient sensing coverage of the monitored area. Our empirical study provides useful insights on how to design distributed routing and in-network processing algorithms tailored for ocean-monitoring sensor networks.

Energy optimization of wireless sensor networks for environmental measurements

In this paper we propose an approach to op t im al design of application-specific wireless sen so r networks based on th e optim ization prop erties of genetic algorithm s. S p eci fi c requirem en ts fo r a precision agriculture applicatio n of sen so r networks are taken in t o a cco u nt by the genetic algo rithm system , together with connectivity an d en erg y co n s erva t i on lim ita tions. We d evel o p an appropriate fitness function to inco rporate many aspects of network performance. The design characteristics optim ized by the genetic algorithm system includ e the sta tu s of sen so r nodes (whether they are active or inactive), network clustering with the ch o i ce of app ro p r ia t e cl u st erh ea d s and fina lly the ch o i ce b et w een two signal ran g es fo r the norm al sen so r nodes. Op tim a l sen so r network designs co nstructed by the genetic algo rithm system satisfy al l application-specific req u irem en ts, fulfill th e ex istent connectivity co nstraints and inco rporate en erg y co n s erva t i o n characteristics.

Optimal design of plant lighting system by genetic algorithms

A genetic algorithm technique is developed for the optimal design of a supplemental lighting system for greenhouse crop production. The approach uses the evolutionary parallel search capabilities of genetic algorithms to design the pattern layout of the lamps (luminaires), their mounting heights and their wattages. The total number and the exact positions of luminaires are not predefined (even though possible positions lay on a fixed grid layout), thus the genetic algorithm system has a large degree of freedom in the designing process. The possibilities of mounting heights and luminaire wattages are limited to four different values for each luminaire in this study. A fitness function for the genetic algorithm was developed, taking into account light uniformity, light intensity capability, shading effects of the design, as well as operational and investment costs. The systems designed by the genetic algorithm show improved values of light uniformity and substantial savings without any effect on the light capacity capabilities of the system. Innovative automatically designed systems compare favorably with typical and expert-designed lighting systems.

A memetic algorithm for optimal dynamic design of wireless sensor networks

We present a memetic algorithm that dynamically optimizes the design of a wireless sensor network towards energy conservation and extension of the life span of the network, taking into consideration application-specific requirements, communication constraints and energy consumption of operation and communication tasks of the sensors. The memetic algorithm modifies an already successful genetic algorithm design system and manages to improve its performance. The obtained optimal sensor network designs satisfy all application-specific requirements, fulfill the existing connectivity constraints and incorporate energy conservation characteristics stronger than those of the original genetic algorithm system. Energy management is optimized to guarantee maximum life span of the network without lack of the network characteristics that are required by the specific sensing application.

Fault Detection and Diagnosis in Deep-trough Hydroponics using Intelligent Computational Tools

The intelligent computational tools of feedforward neural networks and genetic algorithms are used to develop a real-time detection and diagnosis system of specific mechanical, sensor and plant (biological) failures in a deep-trough hydroponic system. The capabilities of the system are explored and validated. In the process of designing the fault detection neural network model, a new technique for neural network designing and training parameterisation is developed, based on the heuristic optimisation method of genetic algorithms. Sensor and actuator faults are detected and diagnosed in sufficient time that the fault detection model can be applied on-line as a reliable supervisor of the operation of an unattended deep-trough hydroponic system. Biological faults were not detected in general. It seems that the interaction between plants and their root-zone microenvironment is not equally balanced, as the condition of the plants is highly influenced by the conditions in their root zone microenvironment, while these microenvironment conditions (as they are represented by the measurable variables) are not influenced in the same degree by the conditions of the plants. Finally, the genetic algorithm system developed here can be successfully applied to a combinatorial problem such as deciding the best neural network architecture, activation functions and training algorithm for a specific model.

Hydroponics water management using adaptive scheduling with an on-line optimiser

An optimisation-based methodology for irrigation control and nutrient supply is developed using common measurements of greenhouse climate. The process measurement has a long time delay, and a feedforward (FF) control loop based on a model-based estimate of water losses is used. A long feedback loop, by which the FF model is adapted using output error feedback, is the mechanism used to minimise the control error. To read the output error, a drain measuring device, or soil moisture meter, is necessary. The optimisation method used is a general tool developed for real-time application and is capable of optimising linear and non-linear systems. The minimisation algorithm used is based on a variant of the Powell direction set method in multiple dimensions. It compares favourably in speed of convergence and accuracy when compared with linear regressors for linear systems. It is therefore used as a generalised tool embedded in a modern greenhouse management system. The method allows on-site on-line identification of plant water needs. As an added benefit, the method provides information for the creation of crop transpiration models.

Deep learning models for plant disease detection and diagnosis

Abstract In this paper, convolutional neural network models were developed to perform plant disease detection and diagnosis using simple leaves images of healthy and diseased plants, through deep learning methodologies. Training of the models was performed with the use of an open database of 87,848 images, containing 25 different plants in a set of 58 distinct classes of [plant, disease] combinations, including healthy plants. Several model architectures were trained, with the best performance reaching a 99.53% success rate in identifying the corresponding [plant, disease] combination (or healthy plant). The significantly high success rate makes the model a very useful advisory or early warning tool, and an approach that could be further expanded to support an integrated plant disease identification system to operate in real cultivation conditions.

Heuristic optimization methods for motion planning of autonomous agricultural vehicles

In this paper, two heuristic optimization techniques are tested and compared in the application of motion planning for autonomous agricultural vehicles: Simulated Annealing and Genetic Algorithms. Several preliminary experimentations are performed for both algorithms, so that the best neighborhood definitions and algorithm parameters are found. Then, the two tuned algorithms are run extensively, but for no more than 2000 cost function evaluations, as run-time is the critical factor for this application. The comparison of the two algorithms showed that the Simulated Annealing algorithm achieves the better performance and outperforms the Genetic Algorithm. The final optimum found by the Simulated Annealing algorithm is considered to be satisfactory for the specific motion planning application.

Feature extraction for time-series data: An artificial neural network evolutionary training model for the management of mountainous watersheds

The present manuscript is the result of research conducted towards a wider use of artificial neural networks in the management of mountainous water supplies. The novelty lies on the evolutionary clustering of time-series data which are then used for the training and testing of a neural object, applying meta-heuristics in the neural training phase, for the management of water resources and for torrential risk estimation and modelling. It is essentially an attempt towards the development of a more credible forecasting system, exploiting an evolutionary approach used to interpret and model the significance which time-series data pose on the behavior of the aforementioned environmental reserves. The proposed model, designed such as to effectively estimate the average annual water supply for the various mountainous watersheds, accepts as inputs a wide range of meta-data produced via an evolutionary genetic process. The data used for the training and testing of the system refer to certain watersheds spread over the island of Cyprus and span a wide temporal period. The method proposed incorporates an evolutionary process to manipulate the time-series data of the average monthly rainfall recorded by the measuring stations, while the algorithm includes special encoding, initialization, performance evaluation, genetic operations and pattern matching tools for the evolution of the time-series into significantly sampled data.