Vehicle Function Control Using Two-Stage GPT Decision Flow
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle control systems using neural networks face challenges due to unpredictable outputs, overfitting, and difficulty in comprehending decision-making processes, especially in universal and large AI systems, which can lead to unreliable vehicle functionality control.
Innovation Solution
A method utilizing generative pre-trained transformers (GPT) in a two-stage execution process to operate vehicle functionalities. The first GPT generates initial output data based on input targets, which are then used to generate further input data for subsequent GPTs or to directly control vehicle functionalities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a universal and large artificial intelligence model is used to control vehicle functionalities, then the adaptability and versatility of the system are improved, but the predictability and comprehensibility of the outputs deteriorate
Solution Approach 1:
The patent divides the single large AI model into multiple specialized sub-models, each trained for specific vehicle control tasks. This segmentation allows each sub-model to maintain high predictability within its domain while the overall system achieves versatility through the combination of multiple specialized models.
Solution Approach 2:
The patent introduces an intermediary layer between the user input and the final control output, consisting of multiple AI sub-models that process information in stages. This intermediary structure enables better interpretation and understanding of the decision-making process while maintaining system adaptability.
2Adaptability or versatility
If a universal and large artificial intelligence model is used to control vehicle functionalities, then the adaptability and versatility of the system are improved, but the comprehensibility of the decision-making process deteriorates
Solution Approach 1:
By segmenting the AI system into multiple specialized sub-models, each handling specific vehicle functions, the patent makes the decision-making process more comprehensible. Each sub-model's decisions can be individually interpreted and understood, unlike a monolithic large model.
Solution Approach 2:
The patent applies preliminary training to each sub-model on specific task datasets before deployment. This preliminary action ensures that each component understands its specific function clearly, making the overall system's decision-making process more transparent and comprehensible.
3Adaptability or versatility
If machine learning methods are used with high diversity of targets in vehicles, then the adaptability of the system is improved, but the reliability of control functions deteriorates due to false positives and overfitting
Solution Approach 1:
The patent segments the control system into multiple specialized AI sub-models, each optimized for specific vehicle control tasks. This reduces overfitting compared to a single universal model, as each sub-model focuses on a narrower set of patterns and targets, improving reliability while maintaining overall adaptability.
Solution Approach 2:
Each AI sub-model is trained with locally optimized quality on specific task datasets relevant to its function. This local quality approach ensures high reliability for each specific control task while the collection of sub-models provides system-wide adaptability.
Data Source
AI summary
A method for operating a vehicle is provided, in which a target to be achieved by at least one vehicle functionality is ascertained and first input data are generated depending on the target. A first generative pre-trained transformer is operated depending on the first input data in order to generate first output data. Second input data are generated depending on the first output data. The first generative pre-trained transformer is operated depending on second input data in order to generate second output data. Alternatively or additionally, at least one second generative pre-trained transformer is actuated depending on the second input data in order to generate the second output data. Finally, the vehicle functionality is operated depending on the second output data.