Perception Error Re-Verification Under Sensor Degradation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Embedded systems in autonomous vehicles face challenges due to sensor degradation and damage over time, which can lead to decreased performance and invalidation of previous system validations, necessitating costly re-verification and potential system shutdown.
Innovation Solution
The proposed solution involves using synthetic data generation with an updated sensor model to re-validate the system, allowing for local re-training and adaptation to sensor alterations, and implementing a monitoring setup to detect and address sensor deficiencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the system uses traditional validation methods with real sensor data, then the validation is accurate and reliable, but the process is computationally expensive and requires significant external resources
Solution Approach 1:
The patent creates synthetic copies of real sensor data by generating simulated sensor outputs that mimic the statistical properties and characteristics of actual sensor measurements. These synthetic datasets replicate the behavior of degraded sensors without requiring physical sensor degradation, enabling validation testing using computationally efficient simulations rather than expensive real-world data collection and processing
Solution Approach 2:
The patent modifies data parameters by applying transformation functions to synthetic sensor data to simulate various degradation scenarios. By changing parameters such as noise levels, signal-to-noise ratios, and degradation patterns through mathematical transformations, the system can evaluate multiple validation scenarios without requiring physical reconfiguration of sensors or extensive computational training
2Reliability
If the system performs comprehensive re-validation after sensor degradation, then system safety is ensured, but operational downtime and service costs increase
Solution Approach 1:
The patent performs preliminary validation using synthetic data before actual sensor degradation occurs or immediately detects it. By pre-generating synthetic datasets that represent potential degradation states and establishing validation thresholds in advance, the system can quickly assess whether degraded sensors still meet safety requirements without requiring lengthy re-validation processes, thus minimizing operational downtime
Solution Approach 2:
The patent enables the embedded system to perform self-validation by processing synthetic sensor data locally using lightweight computational models. This self-service capability allows the system to autonomously determine whether sensor degradation has occurred and whether re-training is necessary, eliminating the need for extensive external validation resources and reducing both time and cost of service interventions
3Ease of manufacture
If the system uses synthetic data for re-validation, then computational resources are reduced and local re-training is enabled, but validation complexity increases
Solution Approach 1:
The patent divides the validation process into distinct segments: synthetic data generation, data transformation, validation testing, and decision-making. Each segment handles a specific aspect of the validation process independently, allowing the system to manage complexity through modular processing steps rather than attempting to handle all validation aspects simultaneously, thus enabling local implementation with reduced computational burden
Data Source
AI summary
A sensor data evaluation device may include an artificial neural network, configured to receive and process first testing data, and output a first testing data output, wherein the first testing data represent sensor data of a first sensor detecting a sensor input; and one or more processors, configured to receive sensor deficiency information representing a deficiency of a second sensor; generate second testing data, wherein the second testing data are based on the sensor deficiency information; implement the artificial neural network to receive and process the second testing data; and output from the output layer a second testing data output; determine a difference between the first testing data output and the second testing data output; and if the difference is outside of one or more predetermined ranges or does not satisfy one or more predetermined criteria, send a signal representing a result of the determined difference.