AI/ML Offloading Interface Between AP and Modem

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current communication devices face challenges in efficiently offloading AI/ML operations between different hardware components, such as Application Processors (AP) and Modems, due to varying compute capabilities and resource competition.

Innovation Solution

The proposed solution involves an apparatus and method that enable communication between AP and Modem using AT commands to inform and request the offloading of AI/ML operations, leveraging shared memory for data exchange, and utilizing available AI/ML compute capacity to optimize resource utilization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If AI/ML operations are performed locally on each entity (AP or Modem), then processing speed and real-time performance are improved, but device complexity and power consumption increase

Engineering Contradiction:

Improveprocessing speedVSAvoiddevice complexity

Core Design Contradiction:

SpeedVSDevice complexity

Solution Approach 1:

The patent segments AI/ML processing capabilities across multiple entities (AP and Modem) rather than concentrating all processing in one location. Each entity can independently perform AI/ML operations when needed, while also being capable of offloading operations to the other entity, thereby distributing the processing burden and reducing individual device complexity while maintaining overall processing speed.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent creates a universal interface that enables both AP and Modem to function as either service providers or service consumers for AI/ML operations. This multi-functionality allows each entity to adapt its role based on available resources and operational needs, reducing the need for dedicated specialized hardware and thereby reducing device complexity while maintaining processing capabilities.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Productivity

If AI/ML operations are offloaded between entities, then resource utilization is improved, but communication overhead and latency increase

Engineering Contradiction:

Improveresource utilizationVSAvoidcommunication latency

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The patent implements preliminary actions by establishing capability advertisement and availability notification mechanisms in advance. Entities proactively inform others of their AI/ML capabilities and resource availability before offloading is needed, allowing receiving entities to pre-configure processing resources and reduce latency when actual offloading occurs.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces a standardized interface as an intermediary layer between AP and Modem for AI/ML operation offloading. This interface handles the complexity of capability matching, operation translation, and resource coordination, thereby simplifying the offloading process and reducing communication overhead while improving overall resource utilization.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Productivity

If proprietary interfaces are used for AI/ML offloading, then vendor-specific optimization is improved, but interoperability between different vendors' hardware components deteriorates

Engineering Contradiction:

Improvevendor-specific optimizationVSAvoidinteroperability

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

Instead of requiring each vendor to implement proprietary client interfaces, the patent inverts the approach by having each vendor implement a standardized service provider interface. The AP and Modem from different vendors both expose standardized capabilities and can consume services from each other, thereby achieving interoperability while allowing vendors to optimize their service provider implementations.

Inventive Principle:

Principle #13The other way round (Inversion)

Data Source

PatentUS20250045629A1Interface to offload ai/ML operations across entities

Publication Date: 2025.02.06 NOKIA TECHNOLOGIES OY

AI summary

Disclosed are methods, apparatuses and computer program products for: informing a second entity of a compute capability of a first entity for taking over one or more artificial intelligence/machine learning operations from the second entity; receiving, by the first entity, a request from the second entity to take over one of the one or more artificial intelligence/machine learning operations from the second entity; performing, by the first entity, the one of the artificial intelligence/machine learning operations to obtain a result of the one of the one or more artificial intelligence/machine learning operations; and providing, by the first entity to the second entity, the result of the one of the one or more artificial intelligence/machine learning operations.