Towards Energy-efficient Data Collection in Internet of Ships

The maritime industry’s increasing focus on energy-efficient vessel operation has generated a demand for costeffective enhancements. The integration of the Internet of Things (IoT) in maritime operations aims to improve safety, efficiency, and environmental sustainability. However, deployment of IoT devices to vessels faces challenges due to potential frequency spectrum congestion and vessel movement beyond terrestrial network coverage, thus evolving into the Internet of Ships (IoS). Addressing energy efficiency and Carbon Dioxide emissions in sea shipping is critical due to its widespread use. In prior researches, we introduced two energy-efficient IoT data collection methods, namely dynamic monitoring frequency and gas level estimation from the transient. Thus, in this research we discuss the importance of selecting the appropriate data collection approach between the aforementioned methods, based on the specific requirements and characteristics of sensors. While the dynamic monitoring frequency approach is suitable for sensors like temperature sensors that can collect data at the moment it is turned on, sensors with preheating, such as gas sensors, are better suited for predefined interval readings to avoid the complexity of building numerous models. However, the new long short-term memory neural network results indicate that reading every second provides the highest accuracy. For reduced data transmission with slightly lower accuracy, selecting 10 logarithmically spaced values offers the best compromise. This approach results in approximately $2 \%$ poorer performance but reduces data points by $50 \%$, thereby conserving energy whether executed on a microcontroller or sent to the cloud. The ultimate objective is to develop energy-efficient solutions that not only reduce sensor energy consumption but also effectively predict future values.