Vehicle Learning Data Selection Using Traveling Data Diversity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for selecting sensor data for machine learning in unspecified devices lack diversity, leading to inefficient learning due to varying device configurations and techniques.
Innovation Solution
An information processing method that associates sensor data with traveling data, determines the degree of difference using a computation model, and selects data based on a threshold value to increase diversity and efficiency in learning data selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If detection values from a specific error determination device are used to select learning data, then the selection is simplified and efficient, but the diversity of learning data for unspecified devices is reduced
Solution Approach 1:
The patent uses traveling data (location, time, weather, road conditions) that are universally available across different devices and configurations, rather than device-specific detection values. This allows the learning data selection to be both efficient and adaptable to various device types, achieving multi-functionality in the selection system.
Solution Approach 2:
The patent changes the selection criteria from device-specific detection values to environmental and operational parameters (traveling data) that can vary across different contexts. This parameter change enables diverse learning data selection while maintaining systematic efficiency.
2Ease of operation
If device-specific detection values are used for learning data selection, then the selection process is straightforward, but the learning data becomes biased toward specific device configurations
Solution Approach 1:
The patent introduces traveling data as an intermediary that mediates between the sensor data and the learning data selection. This intermediary layer decouples the selection process from device-specific configurations, maintaining simplicity while achieving device independence.
Solution Approach 2:
The patent segments the learning data selection into two independent parts: sensor data and traveling data. By evaluating them separately and combining them through association, the system achieves both operational simplicity and adaptability to different device configurations.
3Adaptability or versatility
If all sensor data is collected for learning, then the diversity of learning data is maximized, but the data processing load and time increase
Solution Approach 1:
The patent applies partial action by selecting only a subset of sensor data that is associated with high-diversity traveling data. Rather than processing all sensor data, it selectively processes data that meets the diversity criterion based on traveling data variation, reducing processing time while maintaining diversity.
Solution Approach 2:
The patent performs preliminary action by first evaluating the diversity of traveling data before selecting associated sensor data. This preliminary assessment of traveling data diversity allows for efficient filtering and selection, avoiding the need to process and evaluate all sensor data comprehensively.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided are an information processing method, an information processing apparatus, and a program that can increase a diversity of learning data for configurations or techniques of unspecified devices. The information processing method includes: obtaining sensor data obtained by a sensor installed in a vehicle (200), and at least one type of traveling data of the vehicle (200); associating the sensor data and the at least one type of traveling data with each other; determining a degree of difference of the at least one type of traveling data from the at least one type of one or more traveling data associated with one or more sensor data; and selecting the sensor data as learning data according to the degree of difference.