A pursuit learning solution to underwater communications with limited mobility agents

Underwater environments are subject to varying conditions which might degrade the quality of communications. In this paper, we propose an adaptive control mechanism to improve the communication in underwater sensor networks using the theory of Learning Automata (LA). Our LA based solution controls the mobility of thermocline sensors to improve the link stability in underwater networks. The problem is modelled as a variant of the Stochastic Point Location (SPL) problem [14, 20, 25]. The sensor is allowed two directions of movement, either surface or dive, in order to avoid physical phenomena that cause faults. Our proposed scheme constitutes also a contribution to the field of LA and particularly to the SPL problem by resorting to the concept of pursuit LA. In fact, pursuit LA exploits more effectively the information from the environment than traditional LA schemes that are myopic and use merely the last feedback from the environment instead of considering the whole history of the feedback. Experimental results show the performance of our algorithm and its ability to find the optimal sensor position.