Sensor Data Modification for Cross-Vehicle ML Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning systems for classifying objects using sensor data from vehicles struggle to accurately adapt to different sensor configurations and mounting locations, leading to inconsistent performance across vehicles with varying sensor characteristics.
Innovation Solution
A method is employed to modify sensor data from a first vehicle system to align with the characteristics of a second vehicle system, using a sensor measurements modifier to adjust data based on specified characteristics such as field-of-view, detection range, resolution, and sampling intervals, enabling effective training of a machine learning system to classify objects accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If sensor data is collected from vehicles with different sensor configurations and mounting locations, then the quantity of training data increases, but the consistency and accuracy of machine learning classification deteriorates
Solution Approach 1:
The patent transforms sensor data by adjusting parameters such as field-of-view angles, detection ranges, resolution levels, and sampling intervals to match reference sensor characteristics. This parameter transformation enables data from diverse sensor configurations to be normalized into a consistent format, allowing increased training data quantity without sacrificing classification accuracy
Solution Approach 2:
The patent introduces a sensor measurements modifier as an intermediary component that acts as a bridge between raw sensor data from various configurations and the machine learning training process. This modifier standardizes the data format and characteristics, enabling consistent processing of heterogeneous sensor data while preserving the benefits of diverse data sources
2Stability of the object's composition
If sensor data is standardized to match reference characteristics, then machine learning training consistency improves, but data adaptability to different sensor configurations deteriorates
Solution Approach 1:
The patent implements a dynamic data transformation process where the sensor measurements modifier adapts its transformation parameters based on the source sensor's characteristics and the target reference characteristics. This dynamic adjustment maintains standardized output format while accommodating various input configurations, thus preserving both consistency and adaptability
3Measurement precision
If sensor measurements are transformed to match reference field-of-view and detection range, then object classification accuracy improves, but processing complexity increases
Solution Approach 1:
The patent performs preliminary transformation of sensor data to match reference characteristics before feeding data into the machine learning training process. By pre-standardizing the data format, field-of-view, and detection range parameters in advance, the system simplifies the overall processing pipeline and reduces computational complexity during training, while maintaining high classification accuracy
Data Source
AI summary
A computer that includes a processor and a memory, the memory including instructions executable by the processor for actuating a component of a device based on a parameter output from a machine learning application trained with first output data from a first sensor that has been (1) modified in accordance with a first specified characteristic of the first sensor, and (2) modified in accordance with a second specified characteristic of a second sensor.