Improved <scp>dDQL</scp>: A Double Deep Q‐Learning Enabled Localization for Internet of Underwater Things
Nellore Kapileswar, Judy Simon, Polasi Phani Kumar, Tom Chen, Mithileysh Sathiyanarayanan
Abstract
ABSTRACT Reliable sensor node localization is essential for internet of underwater things (IoUT) applications because it allows management, communication, and sensing in large, uncharted oceanic environments. This research focuses on developing a learning‐enabled node localization model for IoUT using autonomous underwater vehicles (AUVs). To estimate the locations of AUVs, active and passive sensor nodes, a double deep Q‐learning (dDQL) based localization algorithm is introduced. AUVs serve as mobile anchor nodes, and the algorithm uses an online value iteration process to optimize node locations. Active sensor nodes initiate the localization process by transmitting messages, whereas passive sensor nodes determine their location without sending signals. Furthermore, the proposed algorithm for exaggerated crayfish optimization (ExCo) utilizes the selection of optimal actions. The proposed dDQL with ExCo acquired RMSE, localization error, time, delay, throughput, and energy consumption of 1.44E‐07 m, 7.19E‐08 m, 16153.16 s, 13.08 s, 0.98 bps, and 0.35 J, respectively.