Edge Machine Learning Platform for Real-Time Sensor Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing IoT environments face challenges in processing and analyzing large volumes of high-rate, high-volume streaming data in real-time due to limited computing capacity and the need for local context, while cloud-based solutions lack real-time responsiveness and require substantial bandwidth, leading to delayed decision-making.
Innovation Solution
An edge computing platform with machine learning capability is developed to process sensor data in real-time, adapt models for constrained resources, and facilitate seamless model updating and chaining, using a closed-loop arrangement for continuous evaluation and iteration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If cloud-based computing is used to process sensor data, then computing power and storage capacity are improved, but real-time processing capability deteriorates due to data transmission delays
Solution Approach 1:
The patent segments the computing architecture into edge computing nodes deployed at local facilities and remote cloud data centers. Edge nodes process sensor data locally in real-time, while the cloud handles non-time-critical analytics and model training, resolving the contradiction between computing power and real-time processing speed.
Solution Approach 2:
The patent introduces edge computing infrastructure as an intermediary between sensors and cloud data centers. This intermediary processes data locally before cloud transmission, enabling real-time responses at the edge while maintaining cloud connectivity for comprehensive analytics, thus bridging the gap between local speed and remote power.
2Loss of information
If large volumes of sensor data are transmitted to remote cloud, then data analysis capability is improved, but bandwidth consumption and cost increase
Solution Approach 1:
The patent extracts only essential data elements and processed insights from edge facilities for transmission to the cloud, rather than transmitting all raw sensor data. This selective extraction maintains comprehensive data analysis capability while significantly reducing bandwidth consumption and transmission costs.
Solution Approach 2:
The patent performs preliminary data processing, filtering, and aggregation at edge computing nodes before cloud transmission. This preliminary action reduces the volume of data requiring bandwidth-intensive cloud transmission while preserving the analytical value needed for comprehensive insights.
3Speed
If machine learning models are deployed at edge with constrained resources, then real-time processing is improved, but model accuracy deteriorates
Solution Approach 1:
The patent implements local quality optimization by deploying specialized machine learning models tailored to edge computing constraints. These models use quantization, pruning, and hardware-specific optimizations to achieve real-time processing speeds while maintaining acceptable accuracy levels for local decision-making.
Solution Approach 2:
The patent merges multiple model training locations (cloud and edge) with different accuracy-speed tradeoffs. Complex models are trained in the cloud with full accuracy, then adapted for edge deployment with optimized performance, combining the strengths of both environments to achieve both real-time processing and high accuracy.
4Adaptability or versatility
If heterogeneous sensor environments are supported, then adaptability is improved, but software development complexity increases
Solution Approach 1:
The patent implements a universal data ingestion framework at edge facilities that can handle multiple sensor types, protocols, and formats through standardized interfaces. This multi-functional approach enables support for heterogeneous sensor environments while reducing software development complexity through abstraction and standardization.
Data Source
AI summary
An edge computing platform with machine learning capability is provided between a local network with a plurality of sensors and a remote network. A machine learning model is created and trained in the remote network using aggregated sensor data and deployed to the edge platform. Before being deployed, the model is edge-converted (“edge-ified”) to run optimally with the constrained resources of the edge device and with the same or better level of accuracy. The “edge-ified” model is adapted to operate on continuous streams of sensor data in real-time and produce inferences. The inferences can be used to determine actions to take in the local network without communication to the remote network. A closed-loop arrangement between the edge platform and remote network provides for periodically evaluating and iteratively updating the edge-based model.