Local Deep Learning Server for Multi-Client Sensor Inference

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing edge devices are insufficient for performing multiple machine learning tasks simultaneously due to resource limitations, and deploying models in the cloud raises concerns about privacy, security, data bandwidth, and real-time latency.

Innovation Solution

A local deep learning server (DLS) provides scalable access to multiple machine learning instances, optimizing resource use and enabling efficient preprocessing and inference management within a local network, using a customizable architecture and protocols like gRPC/HTTP/2 to handle multiple client requests.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Power

If machine learning models are deployed in the cloud, then processing power and model capacity are improved, but privacy and security are worsened due to data leaving the local network

Engineering Contradiction:
Improveprocessing powerVSAvoidprivacy and security
Core Design Contradiction:
PowerVSObject-affected harmful factors

Solution Approach 1:

The patent introduces a local server as an intermediary between edge devices and cloud services. This server runs machine learning models locally within the home network, serving as a mediator that enables cloud-like processing capabilities without requiring data to leave the local network, thus maintaining privacy and security while providing enhanced processing power

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If data is sent to the cloud for processing, then model accuracy is improved, but data bandwidth consumption is worsened

Engineering Contradiction:
Improvemodel accuracyVSAvoiddata bandwidth
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts the machine learning model execution from remote cloud servers and places it locally on a server within the home network. This extraction eliminates the need to continuously transmit data to the cloud for processing, thereby maintaining model accuracy while dramatically reducing data bandwidth consumption for model inference

Inventive Principle:
Principle #2Taking out (Extraction)

3Power

If data is transmitted to external services, then processing capability is improved, but real-time latency is worsened due to network dependency

Engineering Contradiction:
Improveprocessing capabilityVSAvoidreal-time latency
Core Design Contradiction:
PowerVSLoss of time

Solution Approach 1:

The patent shifts the processing dimension from external cloud-based services to an internal local network infrastructure. By deploying machine learning models on a local server within the home network, the system eliminates network transmission delays while maintaining enhanced processing capabilities, thus achieving real-time performance

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

4Adaptability or versatility

If multiple machine learning tasks are performed on edge devices, then task diversity is improved, but resource sufficiency is worsened

Engineering Contradiction:
Improvetask diversityVSAvoidresource sufficiency
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The patent segments the computational workload by separating edge devices from the machine learning model execution environment. Edge devices continue to perform their native functions while a dedicated local server handles machine learning tasks, allowing multiple diverse tasks to be performed simultaneously without overloading individual edge device resources

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3924845B1Responding to machine learning requests from multiple clients
Publication Date: 2025.09.03 HEWLETT PACKARD DEVELOPMENT COMPANY LP
  • EP3924845B1 patent drawingFigure 1
  • EP3924845B1 patent drawingFigure 2
  • EP3924845B1 patent drawingFigure 3

AI summary

A method includes receiving, with a computing device, a first client request from a first client that identifies a machine learning model and a sensor. The method includes sending, with the computing device, a call to a server to apply the identified machine learning model to a set of data from the identified sensor, in response to the first client request. The method includes receiving, with the computing device, a second client request from a second client that identifies a same machine learning model and sensor as the first client request. The method includes sending, with the computing device, response data from the identified machine learning model to both the first client and the second client without sending an additional call to the server in response to the second client request.