Optimization Technique for Indoor Localization: A Multi-Objective Approach to Sampling Time and Error Rate Trade-off

Magnetic fingerprint-based indoor localization systems are a promising alternative to GPS in indoor environments by using magnetic fields (MFs) to determine accurate indoor locations. However, the optimal number of MFs required for precise location prediction on the client side (CS), i.e., mobile phone, remains a critical yet unanswered question. Hence, resource-constrained devices face limitations in terms of computational power and energy supply. Therefore, this study aims to optimize prediction error and prediction time at CS with minimal MF samples during the online location prediction phase. This study tackles this challenge by employing efficient clustering techniques and machine learning classifiers to minimize prediction error and prediction time on client devices using multi-object optimization instead of single-objective optimization. Our research findings reveal an optimal sample size of 14 per second that provides an ideal balance, offering precise accuracy while reducing the computational burden and energy consumption, thus contributing to the advancement of efficient indoor positioning systems.