Sensor Data Fusion With AI Linking and Entropy-Based Curation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor data fusion systems fail to create actionable data by correlating and fusing heterogeneous, partially heterogeneous, or homogeneous data sources efficiently, leading to excessive computational and storage requirements, and they do not generate new datasets that enhance sensor accuracy or predict future events.
Innovation Solution
A system and method for sensor data fusion that includes a computer processor with curation, linking, fusion, inference, and validation engines, using artificial intelligence to mathematically link sensor outputs exceeding a threshold, creating a unique dataset that minimizes power consumption and storage demands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data from multiple sources is fused to create actionable data, then sensor accuracy and predictive capability are enhanced, but computational and storage requirements increase excessively
Solution Approach 1:
The patent segments the data fusion process into distinct functional engines: curation engine for data collection and filtering, linking engine for establishing relationships, fusion engine for combining data, inference engine for generating insights, and validation engine for verifying results. This segmentation allows each component to process only relevant portions of data independently, reducing overall computational complexity while maintaining fusion benefits
Solution Approach 2:
The curation engine performs preliminary actions by collecting, filtering, and organizing sensor data before it enters the fusion process. By pre-processing and curating data in advance, the system reduces the volume of raw data that requires intensive computational processing in later stages, thereby lowering computational requirements while preserving data quality for accurate sensor fusion
2Loss of information
If heterogeneous sensor data is correlated and fused to generate new datasets, then actionable insights are created, but storage demands increase
Solution Approach 1:
The fusion engine extracts only the essential correlated features and relationships from heterogeneous sensor data to create condensed new datasets. Instead of storing all raw sensor data, the system extracts key actionable insights and stores only these distilled representations, significantly reducing storage requirements while preserving the valuable actionable information
Solution Approach 2:
The validation engine discards redundant or low-value data that does not contribute to actionable insights, while recovering and preserving only the essential correlated information. This selective discarding and recovery process reduces storage demands by eliminating unnecessary data while maintaining the core actionable datasets needed for decision-making
3Reliability
If mathematical linking of sensor outputs is performed to validate accuracy, then data reliability is improved, but processing time increases
Solution Approach 1:
The validation engine performs partial validation by applying mathematical linking only to critical data pairs that have strong correlation potential or high impact on decision-making. Instead of exhaustively validating all possible sensor combinations, the system focuses computational effort on the most significant relationships, reducing processing time while maintaining data reliability for key measurements
Solution Approach 2:
The system applies different validation intensities to different data sources based on their local characteristics and reliability requirements. High-criticality sensors receive more rigorous mathematical validation, while lower-criticality sensors undergo lighter validation processes. This localized quality approach ensures data reliability where needed most while minimizing overall processing time
Data Source
AI summary
Systems and methods of sensor data fusion including sensor data capture, curation, linking, fusion, inference, and validation. The systems and methods described herein reduce computational demand and processing time by curating data and calculating conditional entropy. The system is operable to fuse data from a plurality of sensor types. A computer processor optionally stores fused sensor data that the system validates above a mathematical threshold.