Dynamic Neural Net Configuration for Automated Vehicle Guidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fully automated motor vehicle guidance systems face challenges in adapting to multiple driving situation classes due to the complexity of the physical world, as conventional decision algorithms and neural nets can only be trained for specific scenarios, making it impossible to account for all possible driving situations, and require extensive computing power to implement a high number of individual analysis functions.
Innovation Solution
The system dynamically configures and activates neural nets during runtime based on current driving situation classes, using configuration parameter sets stored in a database to produce analysis units that can be hardware or software-based, allowing for flexible allocation of computing resources and continuous learning through error datasets from a central processing unit.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional decision algorithms and neural nets are trained for specific scenarios, then the system can achieve reliable decisions for those scenarios, but it becomes impossible to account for all possible driving situations
Solution Approach 1:
The system dynamically configures and activates different neural nets during runtime based on the current driving situation class. Instead of using a fixed, static neural net architecture, the system adapts its computational structure in real-time by selecting and configuring appropriate neural nets from a pool of pre-trained models, allowing it to handle diverse driving situations while maintaining reliability for each specific scenario
Solution Approach 2:
The system divides the overall decision-making process into multiple specialized neural nets, each trained for specific driving situation classes. By segmenting the complex task of handling all possible driving situations into smaller, specialized sub-tasks managed by individual neural nets, the system achieves high reliability for each scenario while maintaining overall versatility through the collection of specialized models
2Adaptability or versatility
If a high number of individual analysis functions are implemented to cover all driving situations, then the system can handle diverse scenarios, but extensive computing power is required
Solution Approach 1:
The system uses dynamic configuration of neural nets based on the current driving situation class, activating only the necessary analysis functions when needed. This dynamic approach allows the system to maintain high adaptability across diverse driving situations while optimizing computing power usage by avoiding the continuous execution of all possible analysis functions simultaneously
Solution Approach 2:
The system implements a pool of pre-trained neural nets that cover more driving situations than can be handled simultaneously. By having more analysis functions available than needed at any given moment, the system ensures comprehensive coverage of all possible driving situations while using only the necessary subset during each specific operation, thereby managing computing power requirements
Data Source
AI summary
The invention relates to a method for operating a motor vehicle system which is designed to guide the motor vehicle in different driving situation classes in a fully automated manner. The method includes ascertaining a current driving situation class from multiple specified driving situation classes using at least some of driving situation data, each driving situation class is assigned at least one analysis function. The method further includes retrieving configuration parameter sets assigned to the analysis functions of the current driving situation class from a database and producing analysis units which carry out the analysis function and which have not yet been provided by configuring configuration objects using the retrieved configuration parameter sets.