Automated Driving ODD Validation Using Shared Statistical Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated driving systems (ADS) face challenges in ensuring safe operation within their designed operational design domain (ODD) due to the need for extensive data collection and validation, which is computationally and communicatively inefficient, and the risk of prolonged exposure to inappropriate conditions.
Innovation Solution
A method and system that utilize a scenario identifier module to locally update statistical models based on vehicle operations and share these updates among nearby vehicles, allowing for a combined statistical model comparison to determine deviations from the baseline model, triggering appropriate actions if the ODD is exceeded.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all sensor data is sent to cloud for offline processing and analysis, then comprehensive monitoring and validation can be achieved, but data transmission requirements and communication bandwidth become prohibitively large
Solution Approach 1:
The patent extracts only the essential validation information (statistical model parameters and ODD compliance metrics) from the complete sensor data, transmitting only these condensed parameters to the cloud rather than all raw sensor data. This extraction principle reduces communication bandwidth requirements while maintaining the ability to perform comprehensive safety validation.
Solution Approach 2:
The patent introduces an on-vehicle statistical model and ODD monitoring module as an intermediary that pre-processes sensor data locally before cloud transmission. This intermediary filters and condenses raw sensor data into essential statistical parameters, reducing the data volume transmitted to the cloud while preserving the information needed for safety validation.
2Reliability
If extensive data collection is performed to ensure statistical proofs for safety, then ADS risk reduction is achieved, but the time and resources required become intractable
Solution Approach 1:
The patent performs preliminary statistical model updates and ODD compliance checks on-vehicle before cloud validation. By pre-processing data locally and maintaining updated statistical models of the operational design domain, the system reduces the amount of time needed for post-deployment validation and can quickly assess whether the ADS is operating within its verified parameters.
Solution Approach 2:
The patent implements continuous on-vehicle monitoring of ODD compliance using updated statistical models, rather than relying on periodic extensive data collection campaigns. This continuous monitoring approach maintains statistical proofs for safety in real-time, eliminating long data collection periods while ensuring ongoing safety validation.
3Device complexity
If ADS operates without real-time ODD monitoring, then system complexity is reduced, but the risk of prolonged exposure to inappropriate conditions increases
Solution Approach 1:
The patent implements self-service ODD monitoring where the ADS system automatically tracks its own operational parameters against the statistical model of the operational design domain. The system autonomously determines whether it is operating within its verified parameters and triggers appropriate responses, eliminating the need for complex external monitoring infrastructure while reducing risk through real-time self-validation.
Solution Approach 2:
The patent establishes a feedback loop where the statistical model continuously compares actual operational data against the ODD definition, and automatically triggers alerts or safety responses when deviations are detected. This real-time feedback mechanism provides simple yet effective monitoring that quickly identifies and responds to ODD violations, reducing risk without requiring complex monitoring systems.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure relates to methods and system for managing an Automated Driving System of a vehicle. Moreover, the disclosure presents a use of statistical modelling that is utilized to determine if the conditions and/or requirements of the operational design domain (ODD) are met, whereby the ADS validity may be determined. In more detail, the present disclosure relates to the utilization of nearby vehicles' ADSs to reciprocally share their locally updated statistical models. If the statistical models, from the two ADS, collectively show that the baseline statistical model is not valid at this point in time the appropriate actions from each ADS can be taken.