Critical Location Mapping for Targeted ADS Shadow Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing shadow mode testing for autonomous driving systems in production vehicles is inefficient and resource-intensive, particularly when evaluating multiple candidate software simultaneously, as it often fails to expose the software to crucial and challenging driving situations, and route planning based on historical data may not adequately identify critical locations for testing.
Innovation Solution
A locations mapping system that identifies and maps critical and challenging locations by analyzing past vehicle situations, extracting geo-tagged data from accidentology and vehicle performance databases, and matching these locations with similar road segments in digital maps, enabling a scaled-up mapping for targeted shadow mode testing and real-time vehicle rerouting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If shadow mode testing is run continuously in production vehicles, then software performance can be evaluated, but computational power and resources are wasted on normal operating conditions that do not expose challenging situations
Solution Approach 1:
The system pre-identifies and stores challenging locations in advance using historical data and accidentology databases. These locations are prepared beforehand and made available for targeted testing, eliminating the need for continuous random shadow mode testing and reducing computational waste during normal operations
Solution Approach 2:
The system changes the operational parameter from continuous shadow mode testing to event-triggered testing based on location identification. By monitoring whether the vehicle is approaching a pre-identified challenging location, the system activates testing only when necessary, optimizing resource utilization
2Reliability
If multiple candidate software are evaluated simultaneously, then comprehensive testing coverage is achieved, but onboard computational power requirements increase significantly
Solution Approach 1:
The system extracts and separates the location identification function from the software evaluation process. By pre-identifying challenging locations independently of the candidate software being tested, the system reduces the computational burden during simultaneous multi-software evaluation while maintaining comprehensive testing coverage
Solution Approach 2:
The testing process is segmented into two independent parts: location identification (done once using historical data) and software evaluation (done continuously on identified locations). This segmentation allows multiple candidate software to be evaluated simultaneously without compounding computational requirements
3Productivity
If route planning is based on historical crucial locations, then testing focuses on relevant scenarios, but the system cannot identify new or emerging challenging situations
Solution Approach 1:
The system incorporates feedback loops where shadow mode testing results and new challenging situations are fed back into the location database. This continuous feedback mechanism allows the system to update and expand its knowledge of challenging locations, enabling it to adapt to new scenarios while maintaining focus on proven critical areas
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure relates to a method performed by a locations mapping system (1) for supporting identifying locations crucial to a vehicle (2) equipped with an Automated Driving System, ADS (21). The locations mapping system obtains (1001) data of crucial locations (3) of past vehicle situations identified as critical and/or challenging; extracts (1002) from the crucial locations data, for at least a first past vehicle situation, static road features and operating conditions pertaining to the at least first past vehicle situation; identifies (1004) road segments (4) , e.g. of an ADS-compliant digital map, exhibiting road features to a predeterminable extent matching the static road features pertaining to the at least first past vehicle situation; and generates (1005) a mapping (40) associating the identified road segments with said operating conditions. The disclosure also relates to a locations mapping system in accordance with the foregoing, an arrangement - for instance an offboard system (112) and/or a vehicle comprising such a locations mapping system, and a respective corresponding computer program product and non-volatile computer readable storage medium.