Multi-Sensor Data Fusion for Reliable Cognitive Decision Support
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solutions for fusing and reconstructing sensor data in artificial cognitive systems do not effectively address the need for increased precision and reliability while handling heterogeneous or temporally misaligned data sources.
Innovation Solution
A decision assistance device and method that utilizes a processing unit with encoding and data fusion units, employing machine learning algorithms like generative adversarial networks, to combine and reconstruct data representations from multiple data sources, including virtual sensors, to enhance precision and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data from multiple heterogeneous sensors are fused to improve information completeness and reduce uncertainty, then measurement precision and reliability are improved, but device complexity and difficulty of processing increase
Solution Approach 1:
The system segments the complex data fusion problem by creating separate processing units for each sensor type (first processing unit for first sensor data, second processing unit for second sensor data). Each processing unit independently processes its respective sensor data through encoding and reconstruction operations, then results are combined in a fusion unit. This segmentation reduces overall system complexity while maintaining precision benefits of multi-sensor fusion.
Solution Approach 2:
The patent introduces intermediate representations (encoded data representations in common representation space) as mediators between raw sensor data and final fused results. These intermediate representations standardize heterogeneous sensor data into a common format, facilitating easier fusion while preserving measurement precision. The encoding units act as intermediaries that transform diverse sensor inputs into unified representations.
2Reliability
If virtual sensors are introduced to reconstruct missing data and improve system reliability, then reliability is improved, but device complexity and computational requirements increase
Solution Approach 1:
The system creates virtual sensors that generate reconstructed data representations as copies or substitutes for missing physical sensor data. When a sensor fails or provides incomplete data, the corresponding processing unit uses encoding and reconstruction operations to generate a virtual copy of the missing information based on available sensor data and learned patterns. This copying approach improves reliability without requiring additional physical sensors.
Solution Approach 2:
The patent implements preliminary training phases where the system learns normal operating patterns and relationships between sensors before actual operation. During this preliminary action phase, the system is trained on complete sensor data sets to establish baseline representations and reconstruction models. This pre-learning enables the system to reliably reconstruct missing data during operation, improving reliability while managing complexity through pre-computed models.
3Measurement precision
If machine learning algorithms are used during training phase to learn from complete data sets, then measurement precision and data quality are improved, but loss of time and computational resources increase
Solution Approach 1:
The system performs comprehensive machine learning training and model learning during a preliminary training phase using complete data sets from all sensors. This preliminary action consolidates the time-consuming learning process before deployment, allowing the system to achieve high measurement precision through thorough training. Once trained, the model can operate efficiently with reduced computational requirements during actual cognitive operations, balancing training time against operational precision.
Data Source
AI summary
A device for decision support of a cognitive system based on data originating from a plurality of data sources, the device including a processing unit associated with each data source, each processing unit including an encoding unit configured to determine, from the data from the data source associated with the processing unit, a representation of data in a common representation space by applying a machine learning algorithm to the data, the device further including a data fusion unit configured to determine a representation model of an environment of the cognitive system by combining the data representations determined by the encoding units associated with the plurality of data sources through the application of a data fusion algorithm.