Abstract
Identifying the number and location of processing machines in underwater Wireless Sensor Networks (UWSNs) is one of the hot topics nowadays. UWSNs are vital in monitoring and detecting objects or phenomenon in underwater environment [11]. UWSNs, however, have some limitations and challenges. The low bandwidth capacity is a key challenge [10] [5]. The next main challenge in UWSNs is having long propagation delay [8] [5]. These two challenges negatively impact the performance of UWSNs even if the number and location of processing machines are chosen optimally. Therefore, in paper, we propose a framework including a Modified K-Medoids algorithm that can help to identify the location of processing machines that we need to deploy. We study the effectiveness of having such algorithm on end to end delay and load balancing. Semi-uniform distribution outperforms in term of load balancing comparing to the other two distributions. We consider three different scenario to show merits of our work.