Warehouse Pick Sequence Generation Using Hierarchical Taxonomy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computer systems face challenges in determining efficient sequences for picking ordered items in a warehouse due to sparse data and the complexity of multidimensional scaling algorithms, which require large amounts of data and computational resources, and fail to accurately predict paths when the starting point for picking orders varies.
Innovation Solution
The system reduces complexity by calculating pairwise relations between product categories using historical pick data and applying them to a hierarchical taxonomy, generating sequences that leverage machine learning to infer efficient paths without requiring ordinal information, thus optimizing computation and memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If MDS is used to model physical layout based on time elapsed, then approximate locations can be inferred, but computational complexity and memory requirements become unreasonably large at production scale
Solution Approach 1:
The patent segments the item space into hierarchical clusters (e.g., departments, aisles, sections) rather than treating all items uniformly. This segmentation reduces the computational complexity from O(n²) for pairwise comparisons of all items to O(k²) for comparisons of cluster centroids, where k << n. The hierarchical structure allows efficient route determination by operating at the cluster level while still providing accurate item-level routing.
Solution Approach 2:
The patent introduces cluster centroids as intermediary representations between individual items and the routing algorithm. Instead of directly computing relationships between all pairs of items, the system computes relationships between cluster centroids, which serve as mediators. This intermediary layer dramatically reduces memory requirements and computational load while preserving the essential spatial relationships needed for efficient routing.
2Measurement precision
If MDS requires immensely large amount of data to infer approximate locations, then location accuracy improves, but the approach fails for scenarios with sparse data
Solution Approach 1:
The patent creates a simplified copy or representation of the physical layout using hierarchical clusters and their centroids. This abstracted copy captures the essential spatial relationships and access patterns without requiring complete detailed data about every item's exact position. The cluster-based model serves as a sufficient copy for routing purposes, enabling efficient path determination even with limited observational data.
3Measurement precision
If MDS is used to predict physical layout, then location information can be obtained, but the predictions become inaccurate, ambiguous, or impossible to interpret when starting point varies
Solution Approach 1:
The patent implements a dynamic routing system where the hierarchical cluster structure and route sequences are recomputed based on the actual starting location for each picking task. Rather than relying on a fixed MDS layout that becomes ambiguous with varying starting points, the system adapts the route determination to the specific starting position, maintaining accuracy and interpretability across different scenarios.
4Productivity
If hierarchical product categories are used to reduce complexity, then computation time decreases, but the system must process and store pairwise relations between categories
Solution Approach 1:
The patent transitions from a flat item-level representation to a hierarchical multi-dimensional structure where items are organized into clusters at multiple levels (department, aisle, section). This dimensional change allows the system to operate at the cluster level for routing decisions, reducing the effective data size from millions of items to thousands of clusters, while still providing item-level routing precision when needed.
Data Source
AI summary
An online system performs a method. The method comprises obtaining historical pick data for items located in a warehouse, including data for each of the items picked and pick times between each of the items picked, and determining a taxonomy of items offered by the warehouse. The taxonomy identifies a plurality of product categories structured in a hierarchy, wherein each level of the hierarchy corresponds to a particular level of granularity of product data. The method further comprises applying the historical pick data to a machine learning model to generate pairwise relations between product categories at each level of the taxonomy and generating sequences of product categories based on the pairwise relations. An order for items offered by the warehouse is received and compared to the sequences for each level to generate a pick sequence for picking the items efficiently, which is outputted by the system to a mobile application.