Collaborating Agentic Managers for Low-Latency Edge AI

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Agentic AI systems face challenges in being deployed on edge devices due to memory and computing resource limitations, leading to latency issues and privacy concerns when using cloud-based systems for real-time tasks.

Innovation Solution

A method of collaborating agentic managers on edge devices to process user requests using local AI models, enabling decentralized decision-making and data processing, with features like prompt summarization and runtime app information updates.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If cloud-based systems are used for agentic AI operations, then computing resources and memory capacity are improved, but latency and real-time responsiveness deteriorate

Engineering Contradiction:

Improvecomputing resourcesVSAvoidlatency

Core Design Contradiction:

Quantity of substanceVSLoss of time

Solution Approach 1:

The patent segments the agentic AI system into distributed components deployed across multiple edge devices rather than centralized cloud infrastructure. Each edge device runs local AI models and agentic managers that can independently process requests, eliminating network latency while maintaining distributed computing capabilities through inter-device collaboration.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces device-to-device communication protocols and data sharing mechanisms as intermediaries that enable edge devices to collaborate locally without cloud mediation. This allows real-time information exchange and coordinated decision-making between devices while avoiding the latency inherent in cloud-based architectures.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Power

If cloud-based systems are used for agentic AI operations, then processing power is improved, but data security and privacy deteriorate

Engineering Contradiction:

Improveprocessing powerVSAvoiddata security concerns

Core Design Contradiction:

PowerVSObject-affected harmful factors

Solution Approach 1:

The patent extracts sensitive data processing operations from centralized cloud environments and relocates them to distributed edge devices. This extraction ensures that personal and sensitive data remain on user-controlled devices, eliminating the security risks associated with cloud transmission and storage while maintaining processing capabilities through local AI model execution.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent enables edge devices to autonomously process and manage their own data using local AI models and agentic managers. Each device serves itself by performing inference, decision-making, and data management operations locally, eliminating the need to transmit sensitive information to external cloud servers and thereby ensuring data privacy and security.

Inventive Principle:

Principle #25Self-service

3Speed

If edge devices are used for agentic AI operations, then real-time responsiveness and privacy are improved, but memory size and computing resources deteriorate

Engineering Contradiction:

Improvereal-time responsivenessVSAvoidmemory size

Core Design Contradiction:

SpeedVSQuantity of substance

Solution Approach 1:

The patent implements dynamic model loading and unloading mechanisms that allow edge devices to adapt their memory usage based on current computational needs. AI models can be loaded into memory when required for specific tasks and unloaded when no longer needed, enabling real-time responsiveness while managing limited device memory resources efficiently through dynamic allocation.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent employs model quantization and compression techniques that change the parameter representation of AI models to reduce their memory footprint. By transforming full-precision model parameters into lower-precision representations, the system achieves real-time inference capabilities on resource-constrained edge devices without sacrificing significant accuracy.

Inventive Principle:

Principle #35Parameter changes

4Productivity

If multiple agentic managers collaborate across devices, then task distribution and processing efficiency are improved, but system complexity deteriorate

Engineering Contradiction:

Improveprocessing efficiencyVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent implements a universal agentic manager architecture with standardized interfaces and communication protocols that can operate across different device types and platforms. This universal design enables multiple agentic managers to collaborate efficiently through consistent interaction patterns, reducing the complexity that would otherwise arise from device-specific implementations while maintaining high processing efficiency through distributed task execution.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20260056781A1Framework for Providing Agentic Experiences

Publication Date: 2026.02.26 MEDIATEK INC

AI summary

A first agentic manager and a second agentic manager collaborate to respond to user requests. The first agentic manager on an originating device receives a user request and processes the user request using a first artificial intelligence (AI) model to generate a sequence of sub-requests. The first agentic manager sends one or more of the sub-requests to a second agentic manager on a target device. The second agentic manager processes the one or more of the sub-requests using a second AI model on the target device, and sends an output of an app on the target device to the first agentic manager. Based on the output of the app, the first agentic manager generates a response to the user request on the originating device.