An improved gbln-pso algorithm for indoor localization problem in wireless sensor network / Muhammad Shahkhir Bin Mozamir

Faculty of Computing

Thesis (Master of Science) -- Universiti Malaysia Pahang – 2022

Includes bibliographical references

Wireless Sensor Network (WSN) has become an important field of research. WSN consists of a group of wireless nodes connected between an anchor and unknown nodes. These wireless nodes have the capability to sense the surroundings, process the information and communicate with other nodes wirelessly. The challenging matter in WSN is to estimate the position of the unknown nodes, where there is the error in the distance calculation between nodes. The error in distance estimation phase, caused by noise in range measurement, effects the process of node location. Therefore, the best technique of localization to measure the position of unknown node is required. This study aims to increase the accuracy of node estimation and to minimize time taken for the node localization process. To achieve the stated aims, we implemented an Improved Global best Local Neigborhood Particle Swarm Optimization (IGbLN-PSO) algorithm. IGbLN-PSO algorithm, which is originally from GbLN-PSO algorithm, was applied in previous research into the object tracking problem and it was proved can gain high accuracy and lower the computational time. However, GbLN-PSO searching mechanism must be enhanced when applied into localization problem. This is because the neighbor particles keep searching in the same search space along the main particle’s journey without calculating the optimum value around main particles. This makes the particle calculate the same value, and it may become trapped, while there is the possibility of optimum value around the main particle. Thus, we improved GbLN-PSO, known as IGbLN-PSO algorithm, where the neighbor particles are distributed around the main particle in every iteration to localize unknown node positions. Then, we compared the result with Particle Swarm Optimization (PSO), Differential Evolution Particle Swarm Optimization (DEPSO), Health Particle Swarm Optimization (HPSO) and Global best Local Neigborhood Particle Swarm Optimization (GbLN-PSO) algorithm. The experiment is set to localize forty (40) unknown nodes in 100 × 100 meter area. Three anchors were implemented and the experiments have shown that the accuracy result is competitive where IGbLN-PSO increased 0.3% and 1.5% compared to GbLN-PSO and others, respectively. For result computational time, IGbLN-PSO recorded an increased of 88.88%, 90.99%, 89.75% and 20.49% compared to PSO, DEPSO, HPSO and GbLN-PSO, respectively