Tag-Based Inference Data Routing for Sensor Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In large-scale AI model inference scenarios, especially with numerous sensors and complex data processing, existing methods face challenges in efficiently delivering relevant sensor data to the appropriate inference models, leading to performance degradation due to unnecessary data transmission and lack of applicability determination based on sensor positioning.
Innovation Solution
A computer-implemented method that processes sensor data, generates data and model tags based on characteristics, and uses a time-based tag queue to transmit only relevant data to the appropriate inference models, optimizing data delivery and reducing unnecessary data transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all sensor data is transmitted to all inference models based on full connection practice, then complete data availability is achieved, but system complexity and unnecessary data transmission increase significantly
Solution Approach 1:
The patent segments the full connection architecture by introducing intermediate tag queues that divide data routing into specialized pathways. Data tags and model tags are separated into distinct queues, allowing selective matching rather than universal transmission. This segmentation reduces the complexity of data distribution while maintaining reliability through organized data flow management.
Solution Approach 2:
The patent introduces tag queues as intermediary structures between sensors and inference models. These queues act as mediators that receive data, generate appropriate tags, and selectively route data to matching models. This intermediary layer eliminates the need for direct full connections, reducing system complexity while preserving data availability through structured mediation.
2Reliability
If all sensor data is transmitted to all inference models, then data completeness is maintained, but processing time and computational burden increase
Solution Approach 1:
The patent extracts only the relevant portions of sensor data by generating data tags that identify meaningful characteristics. Instead of transmitting complete datasets to all models, the system extracts and transmits only the tagged data portions that match specific model requirements. This extraction process maintains data completeness for relevant models while significantly reducing processing time and computational burden.
Solution Approach 2:
The patent applies partial action by transmitting data selectively rather than universally. The tag-based routing system determines which inference models receive which data portions, ensuring that each model receives sufficient data for its specific purpose without the excessive action of sending all data to all models. This partial transmission approach reduces processing time while maintaining necessary data completeness.
3Adaptability or versatility
If sensor data is transmitted to all inference models, then data accessibility is maximized, but network bandwidth and energy consumption increase
Solution Approach 1:
The patent segments data transmission pathways by creating separate tag queues for different data characteristics and model requirements. This segmentation enables selective data routing where only relevant data portions are transmitted to specific models, maintaining data accessibility for needed information while reducing overall network bandwidth consumption and energy usage compared to universal transmission.
Solution Approach 2:
The patent applies local quality by tailoring data transmission to specific model requirements through tag-based matching. Each inference model receives data with characteristics locally optimized for its specific purpose, rather than receiving all possible data types. This localized data quality approach maintains accessibility for relevant data while reducing energy consumption by eliminating unnecessary transmissions.
4Reliability
If full connection between sensors and inference models is implemented, then data coverage is complete, but determining sensor data applicability becomes difficult
Solution Approach 1:
The patent introduces tag queues as intermediary structures that solve the applicability determination problem. Data tags generated from sensor data and model tags defining model requirements are matched through the queue system, automatically determining which data is applicable to which model. This intermediary tagging mechanism maintains complete data coverage while eliminating the difficulty of manually determining applicability in full connection architectures.
Data Source
AI summary
An approach to optimize performance for large scale inference models. Data in the form of images is received from sensors such as cameras. The data is processed to generate data tags associated with the context of the image and portion the images. Model tags are generated based on data characteristics or user input. The tags and their associated data are added to a time-based queue for delivery to the appropriate inference models. Based on the embedded delivery time and frequency, the portioned images are delivered to the appropriate inference models.