Scalogram-Based Steering Vibration Detection for Lane Support Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for assessing Lane Support System (LSS) performance in motor vehicles are time-consuming and cumbersome, requiring manual inspection of sensor data to detect steering wheel vibrations and determine reaction time intervals, which can be imprecise and labor-intensive, especially for evaluating the effectiveness of warning vibrations in preventing lane deviations.
Innovation Solution
A method involving a test system that captures temporal signals from steering column sensors, transforms them into event signals, generates scalograms, filters them using thresholds, and uses object detection algorithms to label and extract events, enabling the generation of training data for artificial intelligence to automate the detection of steering wheel vibrations and improve the accuracy of LSS performance assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual inspection of sensor data is used to detect steering wheel vibrations, then detection can be performed, but the process is time-consuming and labor-intensive
Solution Approach 1:
The patent replaces manual mechanical inspection with an automated electronic system that uses signal processing and machine learning algorithms to detect steering wheel vibrations. The system automatically analyzes sensor data, generates scalograms, and identifies vibration events, eliminating the need for manual time-series data monitoring and significantly improving detection productivity while reducing inspection time.
2Measurement precision
If manual inspection of sensor data is used to determine reaction time intervals, then assessment can be performed, but the process is cumbersome and imprecise
Solution Approach 1:
The patent introduces an intermediary automated analysis system that acts as a mediator between raw sensor data and final assessment results. This system uses signal processing techniques (Fast Fourier Transform, scalogram generation) and object detection algorithms to precisely identify vibration start and end points, providing accurate reaction time interval measurements without the complexity of manual inspection procedures.
Solution Approach 2:
The patent transforms the one-dimensional time-series sensor data into a two-dimensional scalogram representation using time-frequency analysis. This dimensional transformation enables more precise detection of vibration events by visualizing frequency content over time, allowing for accurate identification of reaction time intervals that are difficult to detect in raw temporal signals.
3Extent of automation
If automated supervision detection algorithm is used to inspect time series signal, then inspection procedure is automated, but training data generation is required
Solution Approach 1:
The patent applies preliminary action by pre-processing sensor data into scalogram representations and pre-training object detection algorithms with labeled training data before deployment. This preliminary preparation work, including data transformation and model training, enables the system to automatically detect vibration events during actual operation without requiring complex real-time processing, thus achieving high automation while managing system complexity through advance preparation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention describes a method for generating training data (12) for training an artificial intelligence (6) for determining an event (11) in a motor vehicle (1) by a test system (13), comprising the steps of capturing a temporal signal (14) by a capturing device (4) of the test system (13); transforming the temporal signal (14) to an event signal (15) by an electronic computing device (5) of the test system (13); generating a scalogram (16) of the event signal (15) by the electronic computing device (5); providing a first threshold (17) for a magnitude of the event (11) in the event signal (15); filtering the scalogram (16) by using the threshold (17); labeling the event (11) in the filtered scalogram (19) by using an predefined object detection algorithm (21) for the event (11) by the electronic computing device (5); extracting the labeled event (11) from the event signal (15) by the electronic computing device (5); and generating the training data (12) depending on the extracted event (11) by the electronic computing device (5). Furthermore, the invention relates to a method for training an artificial intelligence (6), to a method for determining an event (11) of a motor vehicle (1), to a computer program product, to a test system (13), as well as to an assistance system (2).