Linked Classifiers for Vehicle Surrounding Area State Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Driver assistance systems face challenges in ensuring the correctness and robustness of vehicle surrounding area state determination due to issues with sensor noise and unreliable feature calculations from radar data, particularly for object angles.
Innovation Solution
A method involving two blocks of classifiers, where the first block generates intermediate probabilities from specific sensor data and the second block produces final probabilities, increasing correctness and robustness by using intermediate outputs as inputs, and allowing different classification methods and feature vectors for each block.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning methods are used to determine surrounding area state, then measurement precision is improved, but reliability deteriorates due to difficulty in ensuring correct assessment and robustness against sensor noise
Solution Approach 1:
The system segments the surrounding area into multiple regions (first surrounding area and second surrounding area) and applies different classifiers to each region. This segmentation allows the system to handle different spatial zones with appropriate classification strategies, improving both precision and reliability by tailoring the analysis to specific areas.
Solution Approach 2:
The system introduces an intermediary probability value generated by a first classifier as an additional input feature for the second classifier. This intermediary representation acts as a mediator that captures intermediate assessment results, enabling the final classifier to make more reliable decisions by considering both raw sensor data and processed intermediate probabilities.
2Measurement precision
If model-based methods are replaced with data-based methods, then measurement precision is improved, but device complexity increases due to higher computing capacity demands
Solution Approach 1:
The computing task is segmented into two separate classifier blocks, each handling specific portions of the sensor data. This division distributes the computational load, making the system more manageable and reducing the peak computing capacity requirements while maintaining the benefits of data-based methods.
Solution Approach 2:
The system applies classification only to specific portions of the sensor data rather than processing all data uniformly. The first classifier processes a first portion and the second classifier processes a second portion, with the second classifier receiving both raw data and intermediate probabilities. This partial action approach reduces overall computational complexity while maintaining precision.
3Device complexity
If single classifier system is used, then device complexity is reduced, but measurement precision deteriorates due to inability to handle sensor noise and unknown objects robustly
Solution Approach 1:
The classifier system is segmented into two distinct blocks: a first classifier for initial assessment and a second classifier for final determination. This segmentation enables each classifier to specialize in specific aspects of the task, improving overall precision while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The system introduces an intermediary probability value generated by the first classifier as an additional input feature for the second classifier. This intermediary representation enhances the information available to the final classifier, improving measurement precision by incorporating intermediate assessment results without significantly increasing system complexity.
Data Source
AI summary
A method for determining a state of the surrounding area of a vehicle includes: receiving sensor data of at least one surrounding-area sensor of the vehicle; feeding at least a first portion of the sensor data into at least one first classifier; generating an intermediate probability from the first portion of the sensor data, using the first classifier; feeding at least a second portion of the sensor data and the at least one intermediate probability into a second classifier; generating a final probability of the state of the surrounding area from the second portion of the sensor data and the at least one intermediate probability, using the second classifier.