A Robust Adaptive Algorithm using Inverse Square Root Cost Function for Underwater Channel Estimation

Channel estimation in underwater acoustic (UWA) communication is challenging due to the presence of impulsive noise, extreme multipath spread, and the time-varying nature of the environment. In an impulsive environment, gradient-based algorithms based on the error square cost function fails. Maximum correntropy criteria (MCC), logarithmic cost (LCLMA), and sigmoidal cost functions are used to minimize the effect of impulsive noise with an added cost of convergence rate and complexity. This paper presents a robust cost function, namely Inverse Square Root (ISR), to mitigate the above issue. The slope of the ISR cost function is negligible for high impulsive noise. We compare the effectiveness of the ISR cost function with state-of-the-art cost functions by varying the impulsive noise probability in the UWA channel. In comparison to MCC and LCLMA algorithms, the simulation results show that ISR cost function-based learning converges faster in the presence of a higher percentage of impulsive noise with less computing complexity.