ABSTRACT The integration of software‐defined networking (SDN) with the underwater acoustic sensor network (UASN) has introduced a more centralized and flexible approach to managing underwater sensor nodes. Existing methods of path planning and object detection models have challenges like limited network lifetime and resource constraints that affect efficient path planning and inaccurate object detection. UWSN performance is improved by SDN by providing flexible and centralized management, which can lead to simplified configuration, high energy efficiency, and better adaptability. This advantage makes the proposed model integrate SDN with UWSN; therefore, better energy efficiency can be achieved during path planning and object detection. In UWSN, the traffic flow and network configuration are supported by the SDN control panel. Initially, the nodes are localized in the network scenario using a simplified Kalman filter (SiKaF). Then, the efficient path planning among the AUV and base station is performed using the proposed improved reinforcement learning (ImRL) model. In this, the loss function optimization is employed using the novel adaptive mother optimization (AdMo) algorithm. Finally, using the gathered data, object detection is employed at the base station using the advanced support vector machine (AdSVM) model. AdSVM is designed with skip‐GRU‐based temporal feature extraction and cascaded auto‐encoder–based spatial feature extraction with SVM‐based classification. The extracted spatial and temporal features are concatenated together and fed into the SVM for performing the object detection task. The proposed method acquired an accuracy, IoU, dice coefficient, and MSE of 99.28, 99.14, 0.92, and 0.04, respectively, for the object detection task. During path planning, the energy consumption of the proposed ImRL is 20.573, throughput is 3.022, latency is 2.1 ms, and network lifetime is 21.318.
Vana et al. (Sun,) studied this question.