Agent-Based Evidential Reasoning for EV Adoption Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a lack of historical data for traditional mathematical modeling of electric vehicle adoption, and existing agent-based models are difficult for non-experts to understand due to their complexity and reliance on logical, financial factors, while the purchasing decision for electric vehicles is influenced by subjective factors like image enhancement.
Innovation Solution
An agent-based evidential reasoning decision system using a hierarchical tree structure and agent models to simulate expert knowledge, incorporating multiple factors such as vehicle utility, image enhancement, and financial considerations, to estimate electric vehicle adoption rates at a granular level like zip codes, with outputs visualized on maps for geographic detail.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional mathematical modeling is used, then historical data is required for accurate forecasting, but there is little historic data available for electric vehicle adoption
Solution Approach 1:
The patent introduces an agent-based model as an intermediary between the lack of historical data and the need for accurate forecasting. These agent models represent individual consumers and their decision-making processes, serving as a mediator that can generate adoption forecasts without requiring extensive historical data, thus resolving the contradiction between forecasting accuracy and data availability
Solution Approach 2:
The patent changes the fundamental parameters of the modeling approach by transitioning from traditional mathematical models that rely on historical aggregate data to agent-based models that simulate individual consumer behavior. This parameter change allows the system to forecast adoption rates without requiring the historical data that traditional models would need
2Adaptability or versatility
If complex agent-based models with multiple interacting agents are used, then comprehensive factors can be considered, but the models become difficult for non-experts to understand and credibility is reduced
Solution Approach 1:
The patent segments the complex agent-based model into distinct hierarchical levels: individual consumer agents at the base level, geographic region agents at the intermediate level, and market sector agents at the top level. This segmentation allows comprehensive factor consideration at each level while presenting simplified aggregated results to non-experts, resolving the contradiction between comprehensiveness and understandability
Solution Approach 2:
The patent applies local quality by allowing different levels of the hierarchical model to have different characteristics and complexity. The individual agent level captures detailed subjective factors like image enhancement, while higher levels aggregate these into comprehensible regional and market patterns, enabling non-experts to understand the overall system without being overwhelmed by individual agent complexity
3Device complexity
If models assume agents are driven primarily by logical and financial factors, then the modeling is simplified, but subjective factors like image enhancement that influence purchasing decisions are not captured
Solution Approach 1:
The patent creates a universal agent model framework that can handle multiple types of factors simultaneously - both logical/financial factors and subjective factors like image enhancement. The evidential reasoning algorithm serves as a multi-functional mechanism that processes diverse factor types through a unified structure, resolving the contradiction between modeling simplicity and factor comprehensiveness
Data Source
AI summary
A method and system for an agent-based evidential reasoning decision computer system for determining an adoption rate of a trend is provided. The system includes a plurality of nodes arranged in a tree structure. The plurality of nodes define an evidential reasoning algorithm where lower level nodes receive factors to be considered in the decision and each node assigns a likelihood of an outcome of the received factors, and generates an output to a subsequent higher level node or root of the tree structure. The system also includes a plurality of agent models organized in a hierarchical structure, each agent model comprising a respective set of the plurality of nodes and an output of the agent model, each agent model representing a member of a population, and an aggregator algorithm configured to combine the outputs of the plurality of agent models to generate an output representing an adoption rate.