With advancements and increasing complexity of driver assistance systems, the efforts required for testing, verification and validation also increases. Theoretically each function in the vehicle has to work without failure and finally translate into no accidents until the end-of- life of the vehicle. To ensure this, we need new testing methods.
Billions of kilometers have to be successfully test driven in the real- world before autonomous vehicles are validated and finally released. In practice, this often proves to be difficult, since there is an expo- nential increase in the E/E features and the frequency of software releases is higher than ever before. Covering all tests for each release is time- and cost-intensive. Further main challenges in the industry are frequently changing development, test and validation, Big Data, Real-world complexity and agility in development. AI-Core offers tools and services that allow customer to handle these challenges and make validation efficient and pratical.
AI-Core can process large data and initially extract the ground truth from the existing real data. Valid driving scenarios are classified from the real data and extracted into a database and new test scenarios that focus on critical scenarios are generated.
OUR ADDED VALUE
- High quality scenario classification, extraction directly from the real-world raw data
- Meta information can be added to the available Big Data and add the 5th ‚V‘ which is the Value
- Complies with ASAM standards such as OpenSCENARIO and planned exten sions to upcoming OpenSCENARIO 2.0
- Creation of new scenarios based on system or test requirements or based on criticality with respect to the function under test
- Speeds up the test and validation process
- High quality and quantity annotation and adjusts to changes in the specification with the capability to comply with upcoming standards such as OpenSCENARIO, OpenLABEL, etc.
- Tools and services offer industry
independent solutions for cross
- Data analysis
- HMI testing
- Data Anonymization and GDPR compliance
- Customer specific AI modules to address individual challenges
- Object Annotation and Tracking in camera and LiDAR data
- Seamless scenario based testing
- Scenario classification
- Scenario extraction
- Scenario Clustering
- New scenario generation
- 3D Map generation
- UI:TestAId HMI testing
- Tool for digitalizing classical systems with retrofitment for Industry 4.0
- Video data anonymization
- Result analysis and KPI calculation
- Intelligent Annotaion tool
- Big Data visualization
- AI based Co-driver for real-world testing
- Workshop inspector
- In the near future:
- AI-Core base tool
- Extension to real-world with
automated proving ground
AI-Core is a game changer in the ADAS and autonomous vehicle testing and validation. Resource intensive tasks such as annotation are automated, high quality and quantity results can be achieved. The major breakthrough with AI-Core is the ability to automatically classify and extract dynamic length scenarios from real-world data. These scenarios can replace the manual creation of test scenarios/ test cases which is often complex, error prone and requires high resources. AI provides the ability to learn from the data and adjust to the changing test or system requirements. AI-Core can also cluster/ group unknown scenarios and find similar scenarios which helps to reduce the number of test scenarios.
Since covering all possible scenarios from real-world data is infeasible AI-Core offers a module which can generate new test scenarios for coverage aspects as well as novel scenarios which focus on failures, bugs or critical scenarios tailored to the system under test.
The annotated information, extracted and generated scenarios are stored in databases and can be used for KPI calculations and analysis, replay tests, Xil Tests, etc. Selected or critical scenarios can later be tested in the real-world using automated vehicles in a proving ground.
- Faster, agile, effective and seamless tests are possible for complex systems of ADAS and autonomous vehicles
- Efficient use of real-world big data, making the Big Data usable for test, verification and validation due to high quality and automated processes
- Saves 95% of time in scenario extraction, scenario classification, result analysis and annotation
- Improves quality and at the same time provides deterministic results with real-world Big Data
- Learning algorithms are used and hence reduces the efforts for standardization
- AI-Core supports and allows extensions to automated real-world proving ground extending the closed loop seamless scenario based testing from laboratory to the real-world