The lack of low-cost communication facilities to the shore remains a fundamental problem for fishermen engaged in deep-sea fishing. The Offshore Communication Network (OCN) aims to resolve this connectivity issue by building a wireless network of fishing vessels to provide Internet over the ocean. Effective and reliable wireless communication is crucial for maritime operations to disseminate messages, monitor and track emergency management, and provide information services for enhancing the blue economy. Although OCN shares some of the features of mobile ad-hoc and vehicular networks, they present unique characteristics and research challenges. The primary communication challenges that distinguish OCN from terrestrial networks are (a) the inability to deploy additional infrastructure in the marine environment, (b) the impact of extreme weather conditions on wireless signals, (c) the sea wave-induced movements, (d) the expanded movement freedom in the ocean, compared to the low degree of mobility in terrestrial vehicular networks due to constraints imposed by the road structure and (e) the alignment of directional links between nodes. Providing uninterrupted internet connectivity is a complex problem in OCN for these reasons.
Project Description
This project addresses the critical challenge of improving connectivity among nodes in OCN. A multi-level optimization strategy to solve this connectivity challenge is shown in Figure 1. The research employed a combination of theoretical models, empirical data collection, and simulation techniques to enhance the communication capabilities of OCN nodes.
The study begins with a comprehensive analysis of the factors affecting wireless communication at sea by leveraging real-time data collected from extensive marine experiments on multiple fishing vessels. Based on these inferences, the project proposed a hybrid Bayesian predictive model, as shown in Figure 2, for link quality estimation using historical and real-time data from actual fishing trips. The predictive performance of the learning framework is examined, and it is inferred that the prediction accuracy improved with the hybrid learning scheme compared to other state-of-the-art machine learning models. In addition, a metric to quantify the node’s communication capability to the shore and the peer nodes is proposed. Additionally, the study explored the development of connectivity optimization strategies at three levels, physical, link, and network, to enhance the communication capability of nodes. These algorithms provide better connected node locations, shorter delay packet scheduling, and optimal neighbourhoods for message dissemination. A collection of node position reorientation algorithms is proposed at the physical level to determine suitable locations with better connectivity. These suggestions provide locations with high connectivity at different fishing stages. For link-level optimization, the transmission queue size is estimated for multiple priority messages, and the traffic is scheduled adaptively to minimize queuing delay. Whenever a node is required to transmit a message, it categorizes the traffic according to the context of the communication and places it in the appropriate queue. A multi-priority quasi-birth death model estimates the queue status, and an adaptive RL algorithm is employed for dynamic priority assignment. An adaptive routing scheme that applies reinforcement learning is proposed at the network level to choose the best next hop for improving the packet delivery ratio. The experimental results were interpreted that these three types of methods based on real-time data applied at the physical, link, and network levels improve the communication capability of OCN nodes.
Publication Details
