AI Layout Recommendation for CPQ Configuration Complexity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing configure price quote (CPQ) systems face complexity and maintenance challenges due to combinatorial explosion in configuring highly customizable products, and users desire enhanced visualization of products before purchase.
Innovation Solution
A machine learning model is implemented to recommend layouts for three-dimensional spaces based on historical data, integrating with CPQ systems to suggest structural and non-structural elements, and utilizing extended reality for visualization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a constraint satisfaction engine is used to maintain full set of configuration rules, then the problem of combinatorial explosion is alleviated, but the system becomes complex and difficult to maintain
Solution Approach 1:
The patent replaces the mechanical rule-based constraint satisfaction system with a machine learning model that learns configuration patterns from historical data. Instead of explicitly maintaining and processing configuration rules, the system trains a neural network on historical configure-price-quote data, allowing the model to infer optimal configurations without requiring explicit rule representation, thereby reducing system complexity while maintaining adaptability
Solution Approach 2:
The patent transforms the configuration problem from a rule-based parameter validation approach to a machine learning parameter prediction approach. The system changes the fundamental parameter representation from discrete configuration rules to continuous probability distributions over configuration spaces, enabling the model to handle complexity through statistical learning rather than explicit rule management
2Adaptability or versatility
If compile-based configurator is used to build upon constraint-based engines, then increasingly more complex sets of rules and constraints can be handled, but the system requires compilation of all possible combinations
Solution Approach 1:
The patent performs preliminary action by training the machine learning model on historical data in advance, capturing configuration patterns and relationships before actual product configuration is needed. This pre-learning phase replaces the compile-based approach of pre-computing all possible combinations, allowing the system to handle complex rules without requiring exhaustive compilation of the entire configuration space
Solution Approach 2:
The patent uses historical configure-price-quote data as training examples to teach the model optimal configuration patterns. Instead of compiling all possible configuration combinations, the system learns from copies of historical configurations and their associated price quotes, enabling the model to generalize to new configurations without exhaustive enumeration
3Ease of operation
If users manually configure highly customizable products, then complete control over configuration is achieved, but the complexity of selecting optimal configuration increases rapidly
Solution Approach 1:
The patent enables self-service by having the machine learning model automatically generate configuration recommendations based on historical data patterns. The system serves itself by learning from historical configure-price-quote data and autonomously providing optimal configuration suggestions, reducing the cognitive burden on users while maintaining access to complete configuration control when needed
Data Source
AI summary
Techniques for suggesting a candidate layout based on historical characteristics are disclosed. A system trains a machine learning model to suggest layouts for three-dimensional spaces. The system obtains sets of historical characteristic data, including spatial characteristics of a particular three-dimensional space and layout characteristics including information indicating items present in the particular three-dimensional space and positioning information indicating a position of each item within the particular three-dimensional space. The system trains the machine learning model based on the sets of historical characteristic data. The system receives a first request for a layout suggestion from a first user, including at least a first set of spatial characteristics for a first three-dimensional space. The system applies the machine learning model to the first set of spatial characteristics to identify a first candidate layout as a suggestion and, based on the applying operation: recommends the first candidate layout as a layout suggestion.