Vehicle Control Reliability Check Using Environmental Contribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional vehicle control systems using machine learning models lack assurance that the calculation results are appropriate for actual vehicle control, leading to potential improper control.
Innovation Solution
A vehicle control calculation device that acquires environmental information, specifies important information for control, performs calculations using a trained model, and determines the reliability of the results by comparing important and contribution environmental information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning models are used for vehicle control calculation, then calculation speed and automation are improved, but reliability of control results deteriorates due to inability to guarantee appropriateness of calculation results
Solution Approach 1:
The patent implements a feedback mechanism where the calculated result is fed back into the system to recalculate contribution environmental information. This creates a closed-loop verification process where the system checks whether the environmental information that contributed to the original calculation remains consistent when recalculated from the result, thereby validating the reliability of machine learning-based control calculations without sacrificing calculation speed
Solution Approach 2:
The patent introduces contribution environmental information as an intermediary element that mediates between the machine learning model's black-box calculation and the final control decision. By identifying and tracking which environmental factors contribute most to the calculation result, the system creates a verifiable intermediate representation that bridges the gap between automated calculation and reliable control outcomes
2Measurement precision
If conventional training data improvement techniques are used, then model training quality is improved, but there is no guarantee that actual calculation results will be appropriate for vehicle control
Solution Approach 1:
The patent performs preliminary identification of contribution environmental information before final control decisions are made. By pre-calculating and storing which environmental factors contribute to specific calculation outcomes during the training and validation phases, the system prepares verification data in advance that can be quickly consulted during actual vehicle control operations, ensuring appropriateness without compromising real-time performance
Solution Approach 2:
The system uses feedback loops to continuously verify whether the environmental information identified as contributory during training remains relevant during actual operation. This ongoing verification process ensures that improvements in model training quality translate into appropriately reliable calculation results for actual vehicle control scenarios
Data Source
AI summary
An environmental information acquiring unit for acquiring one or more pieces of environmental information, an important information specifying unit for specifying important environmental information among the pieces of environmental information acquired by the environmental information acquiring unit, a calculation unit for performing a calculation related to control of a vehicle on the basis of the pieces of environmental information acquired by the environmental information acquiring unit, and a trained model in machine learning, a contribution information specifying unit for specifying contribution environmental information among the pieces of environmental information used for the calculation by the calculation unit, and a reliability determination unit for determining a degree of reliability of a result of the calculation performed by the calculation unit by comparing the important environmental information specified by the important information specifying unit with the contribution environmental information specified by the contribution information specifying unit are provided.