Alexandria Engineering Journal (Jan 2025)
TDOA-based WSN localization with hybrid covariance matrix adaptive evolutionary strategy and gradient descent distance techniques
Abstract
Wireless Sensor Networks (WSNs) play a vital role in modern intelligent systems, with localization techniques critical for determining the precise positions of sensor nodes. Time Difference of Arrival (TDOA) is a widely utilized method for node localization. However, its accuracy suffers in noisy environments. This study introduces an innovative hybrid methodology that integrates Covariance Matrix Adaptive Evolutionary Strategy (CMA-ES) with Gradient Descent Distance (GDD) optimization to enhance the accuracy of TDOA-based localization. The hybrid approach utilizes the robust search capabilities of CMA-ES for initial location estimation, thereafter employing GDD to iteratively decrease the mean squared error (MSE) between estimated and real node locations. Simulations performed on several network settings showed substantial improvements. The TDOA-CMAES with GDD strategy demonstrated a maximum enhancement of 24.8 % in localization accuracy, a 36 % decrease in mean localization error (MLE), and an 18 % reduction in computation time relative to traditional approaches such as JAYA and Particle Swarm Optimization (PSO) and other current techniques. The hybrid method produced an Average Localization Error (ALE) much lower than existing techniques, illustrating its resilience in noisy conditions. Our hybrid TDOA-CMAES-GDD localization system is good for real-time WSN applications that need to quickly and accurately locate nodes because it reduces the time it takes to calculate and improves its accuracy. The method thus offers a strong, effective approach for improving TDOA-based localization in noisy settings.
