Pyrih Y. Improving the Efficiency of Wireless Sensor Networks Using a Genetic Algorithm

In the dissertation, a scientific and practical problem of improving the efficiency of wireless sensor network (WSN) operation is solved by developing methods for sensor node placement and for determining an optimal routing path and a set of backup routes based on a genetic algorithm, under conditions of node mobility and heterogeneous communication ranges. The first chapter is devoted to the application of genetic evolution principles in modern networks. It is shown that the genetic algorithm (GA) is one of the most effective tools for optimising network operation processes under dynamic conditions caused by increasing traffic volumes, a growing number of network nodes, and variability in their spatial distribution. It is established that the efficiency of GA application significantly depends on parameter tuning with regard to the problem specifics, network architecture, and computational environment. A classification of genetic algorithms is proposed according to parameter tuning strategies, types of fitness functions, algorithmic variants, and application domains in network-related problems, contributing to a deeper understanding of the factors that determine their performance. The second chapter presents an improved routing method for wireless sensor networks that employs dynamic adaptation of crossover and mutation probabilities based on fitness function values, as well as a set of genetic operators for automatic route reconfiguration in response to topology changes and route length minimisation. Simulation results demonstrate that the proposed GA provides a substantial reduction in route length - by 47.14% compared with the greedy algorithm (GA*) and by 28.39% compared with the ant colony algorithm (ACO) in scenarios with equal node communication ranges - while also reducing the number of hops by factors of three and two, respectively. Thus, the proposed solution exhibits high robustness, adaptability, and optimisation efficiency across all considered WSN topology change scenarios. The third chapter presents a WSN model in a two-dimensional space with nodes characterised by different communication ranges. A multicriteria routing method for WSNs is further developed, employing an evolutionary route search mechanism with a fitness function constructed from normalised network parameters - Euclidean distance, data loss rate, and transmission delay - as well as node characteristics, including battery charge level, signal strength, and node in-degree/out-degree, taking into account their weighting coefficients. This approach enables adaptation of the derived optimal route and the set of backup routes to limited network resources and dynamically changing topologies. To verify the effectiveness of the developed multicriteria GA, it was compared with modified greedy and ant colony algorithms using simulation modelling. The results show that, for a composite criterion combining six individual metrics, the proposed multicriteria GA produces an optimal route that is 15.75% shorter than that obtained using the ant colony algorithm, while the greedy algorithm fails to form a feasible route. Further application of the multicriteria method is considered in the context of a smart city for vehicle routing, taking into account three key criteria: route length, traffic congestion level, and road unavailability. This resulted in a 15.28% reduction in vehicle travel time compared with the ant colony algorithm, whereas the greedy algorithm again failed to reach the destination successfully. These findings confirm the universality of the proposed method and its suitability for both WSN applications and intelligent transportation systems. The fourth chapter addresses the problem of sensor node placement on a plane while minimising the overlap of their coverage areas. For the first time, a spatial placement method for WSN nodes based on a modified genetic algorithm is proposed. The method evaluates the deployment area using node density, penalty mechanisms, and a minimum inter-node distance constraint to determine optimal node configurations during the evolutionary selection process. This enables minimisation of excessive coverage overlap while accounting for heterogeneous communication ranges, both when deploying a new network and when integrating additional nodes into an existing topology. A software module has been implemented in the form of specialised software intended for modelling and simulation-based studies of WSN operation under dynamically changing topologies and limited energy and network resources. Based on simulation results, it is established that the proposed method ensures an effective balance between maximising the number of deployed nodes and minimising their spatial overlap compared with greedy, uniform, and random placement approaches.