ADS Shadow Mode Route Selection for ODD-Based Scenario Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current shadow mode testing for Automated Driving Systems (ADS) in vehicles is inefficient when evaluating multiple candidate software simultaneously, as it requires significant computational resources and often lacks exposure to challenging scenarios, leading to a lengthy development loop.
Innovation Solution
A test-optimizing system that retrieves potential vehicle routes and identifies critical locations within the operational design domain (ODD) of candidate software, allowing for targeted and location-specific shadow mode testing, reducing computational load and increasing the number of software that can be evaluated simultaneously.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If shadow mode testing is performed continuously in production vehicles, then software performance can be evaluated, but computational power and bandwidth are excessively occupied
Solution Approach 1:
The system pre-identifies and stores challenging locations and scenarios in advance within the geographical database, so that during shadow mode testing, the candidate software only needs to be activated when the vehicle enters these pre-defined locations, rather than running continuously. This preliminary preparation reduces real-time computational load while maintaining testing effectiveness
Solution Approach 2:
Instead of running candidate software continuously (excessive action), the system activates it only partially - specifically when the vehicle is at pre-identified challenging locations or encounters relevant scenarios. This partial activation reduces computational power consumption while still gathering sufficient test data for safety evaluation
2Loss of information
If candidate software is activated continuously, then more test data can be collected, but the development loop remains lengthy due to resource constraints
Solution Approach 1:
The system applies different testing strategies to different locations - challenging locations trigger candidate software activation while normal locations do not. This localized quality approach ensures test data is collected where it matters most (at challenging scenarios) without the overhead of continuous activation, thus improving data coverage efficiency and reducing development time
Solution Approach 2:
The system uses the vehicle's own operational data and location information to automatically determine when to activate candidate software, without requiring external continuous monitoring or manual intervention. The geographical database and scenario matching enable the system to self-manage testing activation, reducing resource overhead and accelerating the development loop
3Productivity
If multiple candidate software are evaluated simultaneously, then development efficiency increases, but computational bandwidth becomes insufficient
Solution Approach 1:
The system implements periodic activation of candidate software based on vehicle location and scenario matching, rather than continuous operation. Multiple candidate software can be evaluated simultaneously because each is activated periodically only when relevant challenging scenarios are encountered, reducing cumulative computational bandwidth requirements while maintaining high evaluation throughput
Solution Approach 2:
The system pre-categorizes and stores multiple challenging scenarios and locations in the geographical database, enabling efficient batch evaluation of multiple candidate software. When the vehicle encounters these pre-defined scenarios, multiple software candidates can be activated and evaluated in parallel, increasing productivity without proportionally increasing computational bandwidth consumption
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure relates to a method performed by a test-optimizing system (1) for supporting shadow mode testing of ADS software comprised in an ADS-provided vehicle (2). The test-optimizing system retrieves (1001) one or more potential vehicle routes (3). The test-optimizing system (1) further obtains (1002) for a geographical area covering the one or more potential vehicle routes, data of crucial locations (4) associated with past vehicle situations identified as critical and/or challenging. Moreover, the test-optimizing system retrieves (1003) respective ODD (5) for one or more candidate software (6) respectively adapted to run in the background of the vehicle (2). The test-optimizing system further determines (1004) ODD-compliant locations (7) for respective candidate software, by identifying locations out of the data of crucial locations lying within respective candidate software's ODD. Moreover, the test-optimizing system determines (1005) at least a first test-compliant location (8) along at least a first route out of the one or more potential vehicle routes, by identifying for at least a first candidate software, locations out of the ODD-compliant locations situated along said at least first route. The disclosure also relates to a test-optimizing system in accordance with the foregoing, a vehicle comprising such test-optimizing system, and a respective corresponding computer program product and non-volatile computer readable storage medium.