Abstract
In the last decades, several swarm-based optimization algorithms have emerged in the scientific literature, followed by a massive increase in terms of their fields of application. Most of the studies and comparisons are restricted to high-level languages (such as MATLAB®) and testing methods on classical benchmark mathematical functions. Specifically, the employment of swarm-based methods for solving energy-based acoustic localization problems is still in its inception and has not yet been extensively studied. As such, the present work marks the first comprehensive study of swarm-based optimization algorithms applied to the energy-based acoustic localization problem. To this end, a total of 10 different algorithms were subjected to an extensive set of simulations with the following aims: (1) to compare the algorithms’ convergence performance and recognize novel, promising methods for solving the problem of interest; (2) to validate the importance (in convergence speed) of an intelligent swarm initialization for any swarm-based algorithm; (3) to analyze the methods’ time efficiency when implemented in low-level languages and when executed on embedded processors. The obtained results disclose the high potential of some of the considered swarm-based optimization algorithms for the problem under study, showing that these methods can accurately locate acoustic sources with low latency and bandwidth requirements, making them highly attractive for edge computing paradigms.
Highlights
In the last decade, swarm optimization methods have found their way into the scientific community, where several algorithms have been proposed and applied in real-life problems
We present the execution times of each swarm algorithm on different embedded devices averaged over 10,000 Monte Carlo runs, giving as a reference the execution times of the Grid Search algorithm with a 0.1 m spacing interval
Considering the features of the different algorithms tested and the obtained results, it was shown that when using smart/intelligent initialization, the algorithms that rely more on the local space perform better than the ones with stronger initial exploration phases
Summary
Swarm optimization methods have found their way into the scientific community, where several algorithms have been proposed and applied in real-life problems. Taking advantage of specific information about the problem layout to intelligently initialize the population, an improved EHO (iEHO) showed even better accuracy, with good results over a wide range of measurement noises, network size, and even in tracking scenarios [55,56] This is one of the main reasons for why this work studies swarm-based techniques. Based on the above discussion and the results obtained, the main insights and contributions of the present work are summarized as follows: (1) application of several of the most significant and up-to-date swarm-based techniques to the EBAL problem and assessing their performance with regard to convergence and localization error; (2) integration of the intelligent initialization technique proposed in [55] (but only integrated with EHO).
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
