Supply Chain Resource Allocation Using Precomputed Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing supply chain optimization methods are computationally intractable and inefficient for large-scale networks with multiple destinations and sources, failing to provide optimal resource allocation due to the non-deterministic polynomial-time hardness and large-scale nature of the problem, leading to unacceptable runtime and memory usage.
Innovation Solution
A system that decomposes the supply chain optimization problem into multiple stages, using a combination of simulation, optimization, and heuristics, including Ant Colony Optimization with virtual-node addition and problem inversion, to solve for optimal paths and resource allocation in large-scale networks, allowing for dynamic network adjustments and constraint-based order allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If exact explicit mathematical formulas are used to solve supply chain optimization problems, then optimal paths can be found, but computational complexity becomes intractable for large-scale networks
Solution Approach 1:
The patent segments the large-scale supply chain network into multiple sub-networks or regions, and further divides optimization into two stages: (1) pre-computing and storing optimal paths for all source-destination pairs in each sub-network, and (2) composing these pre-computed paths to solve multi-destination problems. This segmentation reduces the computational complexity from solving the entire large network at once to solving smaller sub-networks independently.
Solution Approach 2:
The patent performs preliminary computation of optimal paths between all source nodes and destination nodes in each sub-network before actual order processing. These pre-computed paths are stored in a data structure that enables efficient querying. When an order arrives, the system quickly retrieves and composes pre-computed paths rather than performing new optimization calculations, significantly reducing real-time computational burden.
2Measurement precision
If separate optimization algorithms are run for each order, then optimal paths can be determined, but runtime becomes unacceptable due to linear scaling with number of orders
Solution Approach 1:
The patent merges multiple individual order optimization problems into a single batch processing operation. By collecting multiple orders and processing them together using the pre-computed path data structure, the system achieves sub-linear scaling relative to the number of orders. The merging allows efficient composition of paths for multiple destinations simultaneously rather than running separate algorithms for each order.
Solution Approach 2:
The system performs preliminary computation of all possible source-destination paths in advance, storing them in an optimized data structure. This preliminary action transforms the problem from performing expensive optimization calculations at order-time to performing simple path composition operations, dramatically reducing runtime per order while maintaining optimal path selection.
3Ease of operation
If graph traversal algorithms are used to find optimal paths, then paths can be determined, but the algorithms do not scale well with multiple edges between nodes
Solution Approach 1:
The patent replaces traditional graph traversal algorithms (like Dijkstra's or Bellman-Ford) with a matrix-based computational approach. By representing the network as adjacency matrices and using matrix operations to compute and store optimal paths, the system achieves better scalability. Matrix operations can be efficiently parallelized and optimized, handling multiple edges between nodes more effectively than sequential graph traversal algorithms.
Data Source
AI summary
Systems and methods for allocating optimal supply chain resources based on order data are disclosed. In some embodiments, a disclosed method includes: receiving, from a computing device, an allocation request associated with an order; determining a customer node associated with the order, wherein the customer node is in a supply chain network; determining, based on the order and the customer node, a sub-network of the supply chain network; determining, based on cost information associated with the sub-network, at least one path from at least one source node to the customer node in the sub-network; generating supply chain allocation data for the order based on the at least one path; and transmitting the supply chain allocation data to the computing device.