GPU Parallel Processing for Stochastic Safety Stock Optimization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current inventory management systems for large multinational companies fail to optimize safety stock levels on a per-item basis for each store, leading to inefficiencies such as overstocking or insufficient inventory, which increases costs and environmental impact due to unnecessary transportation, as they typically rely on fixed demand distributions that do not account for daily variability and seasonality.

Innovation Solution

Implementing a stochastic optimization technique using a specifically programmed graphics processing unit (GPU) to generate sample paths based on historical demand data, considering daily and seasonal variations, to determine optimal safety stock settings that minimize total costs and ensure adequate inventory levels across all stores.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If fixed demand distributions are used for inventory management, then system complexity is reduced, but manufacturing precision of safety stock levels deteriorates

Engineering Contradiction:

Improvesystem complexityVSAvoidsafety stock level precision

Core Design Contradiction:

Device complexityVSManufacturing precision

Solution Approach 1:

The patent changes the parameter of demand distribution from fixed to stochastic, incorporating variability and seasonality parameters to improve safety stock level precision while managing system complexity through structured modeling approaches

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces dynamic demand distributions that adapt to varying conditions including daily variability and seasonality, allowing safety stock calculations to respond to changing demand patterns rather than relying on static fixed distributions

Inventive Principle:

Principle #15Dynamics

2Manufacturing precision

If stochastic optimization with GPU processing is implemented, then manufacturing precision of safety stock levels is improved, but device complexity increases

Engineering Contradiction:

Improvesafety stock level precisionVSAvoidcomputational system complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent replaces traditional CPU-based sequential processing with GPU-based parallel processing, leveraging the architectural differences between these computing devices to handle stochastic optimization computations more efficiently

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent implements Monte Carlo simulation with a specified number of sample paths (e.g., 1000 simulations) to achieve sufficient precision for safety stock calculations without requiring exhaustive computation, balancing accuracy with computational feasibility

Inventive Principle:

Principle #16Partial or excessive action

3Manufacturing precision

If per-item safety stock optimization is performed for all stores, then manufacturing precision is improved, but loss of time in computation increases

Engineering Contradiction:

Improvesafety stock level precisionVSAvoidcomputation time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent performs optimization for a representative subset of store-items (e.g., 5000 store-items out of potentially millions) to demonstrate the methodology and achieve sufficient precision for decision-making without requiring exhaustive computation across the entire enterprise

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The patent divides the large-scale optimization problem into smaller manageable segments, processing store-items in batches or groups, which allows parallel processing and reduces the time required to compute safety stock levels for individual store-items

Inventive Principle:

Principle #1Segmentation

4Ease of operation

If fixed demand distributions are used, then ease of operation is maintained, but productivity of inventory management deteriorates

Engineering Contradiction:

Improvesystem operation simplicityVSAvoidinventory management efficiency

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The patent implements automated stochastic optimization that self-adjusts safety stock levels based on historical demand data and variability patterns, reducing manual intervention while improving inventory management productivity through data-driven decision-making

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS11348047B2Systems and methods for safety stock settings using a parallel processing computing architecture

Publication Date: 2022.05.31 WALMART APOLLO LLC

AI summary

This disclosure describes a graphics processing unit programmed to generate a sample path for the demand of the one or more products at a store based at least in part on data associated with a historical distribution of the variability of the demand of the one or more products. The graphics processing unit may generate a thread corresponding to a plurality scenarios. The graphics processing unit may execute the thread in parallel to determine one or more parameters for each of the plurality of scenarios for the one or more products. The graphics processing unit may select the one or more parameters generated from the execution of one of the sample paths to minimize the cost. The graphics processing unit may adjust an inventory management system to set an inventory management setting based at least in part on the selection of the one or more parameters.