Granted Patent B2

US 12,726,213 · App. 19/192,244 · Granted Sep 1, 2026

Inventors: Joshua Cooper (Columbia, SC); Grant Fickes (Columbia, SC); Charles Yeomans (Orinda, CA)

Assignee: ATOM BEAM TECHNOLOGIES INC.

H03M7/3059G06N20/00H03M7/6005

Abstract

Claims (114)

1 . A computer system comprising:

a hardware memory, wherein the computer system is configured to execute software instructions stored on nontransitory machine-readable storage media that enables dynamic task migration between edge and cloud environments at multiple synchronization points, comprising:

receive raw input data from sensors or other sources;

analyze characteristics of the input data based on a machine learning algorithm;

monitor resource status of edge devices and cloud infrastructure;

evaluate network conditions between the edge devices and cloud infrastructure;

predict future network and resource conditions based on historical performance data;

determine optimal task distribution between the edge devices and cloud infrastructure based on the analyzed characteristics, monitored resource status, evaluated network conditions, and predicted future conditions;

dynamically partition processing tasks between the edge devices and cloud infrastructure according to the determined optimal task distribution;

implement state synchronization between distributed processing components by applying differential encoding to minimize data transfer volume; and

update a performance database with execution metrics to influence future task distribution decisions, wherein a performance feedback loop connects the database updates back to a prediction component to continuous refinement of the system's decision-making capabilities by:

implementing device-specific contextual processing by incorporating location awareness, motion detection, and ambient conditions;

applying transfer learning mechanisms to share optimization knowledge across similar devices; and

providing graceful degradation modes during resource constraints while maintaining essential functionality.

2 . The computer system of claim 1 , wherein the edge devices implement a partitionable processing pipeline comprising:

receiving raw input data from sensors or data sources;

performing initial data cleaning and normalization;

identifying relevant characteristics in the pre-processed data;

converting the data into formats suitable for efficient processing;

performing preliminary evaluation of the transformed data; and

determining actions based on the analysis results.

3 . The computer system of claim 1 , wherein the cloud infrastructure implements a partitionable processing pipeline comprising:

performing computationally intensive feature identification;

implementing data transformations requiring substantial computing resources;

performing analytical operations leveraging cloud-based processing power;

identifying patterns across multiple data sources or historical data; and

applying machine learning models to the processed data.

4 . The computer system of claim 1 , wherein the edge-cloud orchestration subsystem further comprises programming instructions which, when operating on the processor, causes the processor to:

detect a migration trigger indicating a need to redistribute processing tasks;

evaluate whether the migration trigger represents a critical condition requiring immediate action;

establish a migration sequence based on task dependencies;

transfer state information to a target environment;

verify successful transfer of the state information; and

initialize migrated tasks at their new location.

5 . The computer system of claim 1 , wherein the edge-cloud orchestration subsystem further comprises programming instructions which, when operating on the processor, causes the processor to:

collect telemetry data from edge devices;

evaluate connection quality parameters;

examine processing requests to estimate computational requirements;

implement multi-objective optimization for task distribution; and

maintain consistent processing state across distributed components.

6 . The computer system of claim 1 , wherein predicting future network and resource conditions comprises:

analyzing long-term trends using time-series decomposition techniques;

generating medium-term forecasts using ensemble learning methods;

providing short-term predictions using online learning algorithms; and

implementing lightweight prediction models on one or more edge devices.

7 . The computer system of claim 1 , wherein the performance database comprises:

historical execution metrics comprising processing time, energy consumption, and memory usage;

network performance statistics comprising bandwidth, latency, and connection reliability; and

task migration performance data comprising migration time, transfer overhead, and state size.

8 . The computer system of claim 1 , wherein the edge-cloud orchestration subsystem further comprises programming instructions which, when operating on the processor, causes the processor to:

implement adaptive compression based on available bandwidth and network conditions;

apply differential encoding for state synchronization to minimize data transfer volume; and

utilize specialized communication protocols optimized for intermittent connectivity.

9 . The computer system of claim 1 , wherein determining optimal task distribution employs multi-objective optimization that balances:

processing latency minimization;

energy consumption reduction;

bandwidth utilization optimization;

data privacy requirements; and

quality-of-result maximization.

10 . The computer system of claim 1 , wherein the edge-cloud orchestration subsystem further comprises programming instructions which, when operating on the processor, causes the processor to:

provide autonomous operation during network disconnection by implementing:

local decision-making capabilities at the edge;

state caching mechanisms; and

eventual consistency protocols for state reconciliation upon reconnection.

11 . A computer-implemented method for adaptive edge-cloud data processing that enables dynamic task migration between edge and cloud environments at multiple synchronization points, comprising the steps of:

receiving raw input data from sensors or other sources;

analyzing characteristics of the input data based on a machine learning algorithm;

monitoring resource status of edge devices and cloud infrastructure;

evaluating network conditions between the edge devices and cloud infrastructure;

predicting future network and resource conditions based on historical performance data;

determining optimal task distribution between the edge devices and cloud infrastructure based on the analyzed characteristics, monitored resource status, evaluated network conditions, and predicted future conditions;

dynamically partitioning processing tasks between the edge devices and cloud infrastructure according to the determined optimal task distribution;

implementing state synchronization between distributed processing components by applying differential encoding to minimize data transfer volume; and

updating a performance database with execution metrics to influence future task distribution decisions, wherein a performance feedback loop connects the database updates back to a prediction component to continuous refinement of the system's decision-making capabilities;

implementing device-specific contextual processing by incorporating location awareness, motion detection, and ambient conditions;

applying transfer learning mechanisms to share optimization knowledge across similar devices; and

providing graceful degradation modes during resource constraints while maintaining essential functionality.

12 . The computer-implemented method of claim 11 , further comprising the steps of:

detecting a migration trigger indicating a need to redistribute processing tasks;

evaluating whether the migration trigger represents a critical condition requiring immediate action;

identifying task dependencies;

establishing a migration sequence;

creating checkpoints at synchronization points;

transferring state information to a target environment;

verifying successful transfer of the state information; and

initializing migrated tasks at their new location.

13 . The computer-implemented method of claim 11 , wherein predicting future network and resource conditions comprises:

analyzing long-term trends to identify seasonal patterns;

generating medium-term forecasts with confidence intervals; and

providing short-term predictions using adaptive re-training mechanisms.

14 . The computer-implemented method of claim 11 , wherein dynamically partitioning processing tasks comprises:

identifying atomic computational units within a directed acyclic graph of processing modules;

determining optimal execution locations for each computational unit; and

preserving state information at synchronization points to enable task migration.

15 . The computer-implemented method of claim 11 , further comprising the steps of:

implementing priority-based scheduling that distinguishes between time-critical processing functions and background tasks;

applying speculative execution for latency-sensitive applications; and

implementing checkpoint-based recovery mechanisms for resilience during connectivity disruptions.

16 . The computer-implemented method of claim 11 , further comprising the steps of:

collecting performance metrics including processing time, energy consumption, memory usage, and network utilization;

storing the metrics in a time-series database with automatic downsampling; and

applying association rule mining to discover correlations between system conditions and performance outcomes.

17 . The computer-implemented method of claim 11 , wherein monitoring resource status comprises:

implementing a hierarchical monitoring architecture with configurable sampling rates;

supporting both push-based and pull-based metrics collection; and

utilizing statistical process control techniques to identify significant deviations from normal operation.

18 . The computer-implemented method of claim 11 , wherein evaluating network conditions comprises:

employing active probing with variable packet sizes and intervals;

performing passive monitoring of actual application traffic patterns; and

maintaining a dynamically updated connectivity graph with weighted links representing current network conditions.

19 . The computer-implemented method of claim 11 , further comprising the steps of:

implementing context-aware partitioning based on location, application requirements, and user behavior;

adjusting granularity of partitioning based on network stability; and

optimizing for different objectives based on device power source and charging status.

Assignments (1)

Continuity (10)

References Cited (20)