Marine Sensor Data Prioritization for Limited-Link AI Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for collecting and transmitting data from marine vessels to train machine learning algorithms face challenges such as limited communication coverage, quality, latency, and cost, making it difficult to efficiently collect and transmit relevant data for improving navigation safety and reliability.
Innovation Solution
A method and system that prioritize and selectively transmit data based on relevance criteria, using remote communication links when available, allowing for efficient data processing and storage during voyages where communication is not feasible, and immediate transmission when conditions improve, optimizing processing resources and cost.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all sensor data are transmitted continuously, then complete training data is available, but communication cost and bandwidth consumption increase significantly
Solution Approach 1:
The patent applies local quality by differentiating data transmission based on relevance. High-relevance data (e.g., dangerous situations, abnormal events) are transmitted with high priority, while low-relevance routine data are transmitted with lower priority or stored locally. This selective approach ensures critical training data is communicated without wasting bandwidth on redundant information.
Solution Approach 2:
The system changes the parameter of data priority based on relevance assessment. A relevance criterion evaluates each data point and assigns priority levels dynamically. This parameter transformation converts uniform data streams into prioritized transmission queues, optimizing communication resource allocation according to actual training needs.
2Productivity
If data are prioritized and selectively transmitted, then communication efficiency improves, but data selection complexity increases
Solution Approach 1:
The relevance criterion operates autonomously to evaluate and prioritize data without requiring complex external intervention. The system self-assesses data relevance based on predefined criteria (e.g., detection of dangerous situations, abnormal events, or training-specific requirements), automatically assigning priority levels and managing transmission queues.
Solution Approach 2:
Data are prioritized and pre-processed before transmission based on their relevance to training objectives. The system performs preliminary filtering and classification, organizing data into priority queues in advance. This preliminary action reduces the computational burden during actual transmission and simplifies downstream processing.
3Reliability
If data are stored during non-communication zones, then data loss is prevented, but storage requirements and memory usage increase
Solution Approach 1:
The system applies local quality by storing data with different retention policies based on relevance. High-relevance data are preserved in storage memory for later transmission, while low-relevance data may be processed locally or discarded. This selective storage approach maintains critical information without filling memory with redundant data.
Solution Approach 2:
The system discards low-relevance data that do not contribute significantly to training, while recovering and preserving high-relevance data for future transmission. This selective discarding and recovery process optimizes storage utilization by maintaining only the data subset that provides maximum training value.
Data Source
AI summary
A method comprising, by at least one processing unit, obtaining data collected by one or more sensors of at least one marine vessel, said data being representative of one or more situations encountered by the marine vessel during its voyage, prioritizing data according to at least one relevance criterion, wherein when the marine vessel is located in a zone in which at least one remote data communication link meets a criterion, transmitting at least some of the data using the at least one remote data communication link, wherein data are transmitted according to priority determined for the data, thereby facilitating transmission of relevant data for the purpose of training one or more machine learning algorithms (e.g. deep learning algorithms) providing output based on these data.