Edge-Cloud Inference Data Transmission Adaptation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

There is a challenge in determining the optimal amount of information to transmit to a cloud server for inference processing, as excessive transmission increases certainty but wastes capacity, while insufficient transmission reduces inference accuracy.

Innovation Solution

A processing method and system that involves the edge device transmitting first and second data to a server device for inference, with the server requesting additional data only when the initial inference confidence is below a threshold, allowing for adaptive data transmission and improved accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If more information is transmitted to the cloud side, then the certainty of inference result increases, but the transmission capacity is wasted

Engineering Contradiction:

Improvecertainty of inference resultVSAvoidtransmission capacity

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The system dynamically adjusts the amount of information transmitted based on the confidence level of edge inference. When confidence is high, only essential data is sent; when confidence is low, additional data is transmitted. This dynamic adaptation resolves the contradiction by making transmission capacity usage proportional to actual inference needs.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the parameter of information transmission volume based on the confidence level parameter. By adjusting the transmission amount according to inference confidence, the system optimizes both reliability and resource utilization, avoiding both over-transmission and under-transmission.

Inventive Principle:

Principle #35Parameter changes

2Quantity of substance

If less information is transmitted to the cloud side, then the transmission capacity is saved, but the certainty of inference result decreases

Engineering Contradiction:

Improvetransmission capacityVSAvoidcertainty of inference result

Core Design Contradiction:

Quantity of substanceVSReliability

Solution Approach 1:

The system dynamically determines transmission volume based on real-time confidence assessment. This dynamic approach ensures that transmission capacity is conserved when possible while guaranteeing sufficient information is transmitted when needed for accurate inference.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The edge device autonomously assesses its own inference confidence and self-determines the appropriate transmission volume. This self-service mechanism eliminates the need for continuous cloud supervision, enabling efficient local decision-making about resource allocation.

Inventive Principle:

Principle #25Self-service

3Measurement precision

If all data is transmitted to the cloud side for inference, then the inference accuracy is maximized, but the transmission time and energy consumption increase

Engineering Contradiction:

Improveinference accuracyVSAvoidtransmission time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system extracts only the necessary portion of data for transmission to the cloud, leaving sufficient statistics and confidence information at the edge. This extraction principle reduces transmission time and energy while maintaining the cloud's ability to perform accurate inference when needed.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The edge device performs preliminary inference and data preprocessing before transmission, preparing confidence levels and essential features in advance. This preliminary action reduces the amount of data requiring transmission and accelerates the overall inference pipeline.

Inventive Principle:

Principle #10Preliminary action

4Speed

If edge device performs inference independently, then the response speed is fast, but the inference accuracy may be insufficient for complex cases

Engineering Contradiction:

Improveresponse speedVSAvoidinference accuracy

Core Design Contradiction:

SpeedVSMeasurement precision

Solution Approach 1:

The inference task is segmented into two parts: initial inference performed locally at the edge for fast response, and supplementary inference performed at the cloud for enhanced accuracy. This segmentation allows the system to benefit from both speed and precision.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The confidence level acts as an intermediary that bridges edge and cloud inference. It determines when edge inference is sufficient and when cloud inference is needed, enabling seamless coordination between fast local processing and accurate remote processing.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20240338939A1Processing method, and processing system

Publication Date: 2024.10.10 NT T INC

AI summary

A processing system (100) is a processing method for performing inference processing in an edge device (20) and a server device (30), the method including a first transmission process in which the edge device (20) transmits first data based on data to be inferred to a server device that performs first inference, and a second transmission process in which the edge device (20) transmits second data based on the data to be inferred to an execution unit that performs second inference in response to a request from the server device (30), in which the request from the server device (30) is made in a case where a result of the first inference performed in the server device (30) is equal to or less than predetermined confidence.