IoT-based hybrid optimized fuzzy threshold ELM model for localization of elderly persons

• The proposition of the fuzzy logic system (FLS) applied over the centroid and ELM for node localization to handle both the low and the high-density scenarios, respectively. • Designed the control parameter ( k α ) for PSGWO for boosting the decline speed of convergence factor so that local search can be improved and optimization time can be minimized. • Optimized FLS and ELM using PSGWO with a free vector for adjusting approximation precision nearer to the moving node’s actual position. • Proposed a novel population and multi-criteria based soft computing algorithm called hybrid optimized fuzzy threshold extreme learning machine (HOFTELM). Due to the quickly aging population, the number of elderly persons is rapidly increasing, posing significant challenges for monitoring and assisting them in indoor and outdoor settings. Although some techniques are available for the indoor localization of elderly persons, in the coming years, outdoor localization will be an essential part of society. Different approaches such as GPS, range-based, and range-free have been developed for outdoor localization. However, the localization accuracy and precision is still a significant challenge. For accurate and low-cost localization, we propose a novel IoT-based range-based localization for smart city applications. Using the extreme learning machine (ELM), fuzzy system, and modified swarm intelligence, a hybrid optimized fuzzy threshold ELM (HOFTELM) algorithm is developed. The particle swarm grey wolf optimization is used to identify the direction of the moving sensor node. A fuzzy weighted centroid is used to optimize the consequences of irregular movement of the nodes. Lastly, an optimized threshold extreme learning machine and weighted mean are applied to localize the moving nodes accurately. Our algorithm outperforms the existing algorithms in terms of average location error ratio (ALER), the number of localized nodes, and the computational time. The results show that ALER reduces by at least 48.07% in comparison with the other algorithms. The proposed algorithm also localizes at least 7.25% additional nodes and has a computationally efficient operation.