Simulation platform for testing & validation of HAD systems
The AAI ReplicaR platform provides a holistic synthetic world in simulation, combined with a consistent tool chain for virtual validation using closed and open loop testing as well as data-driven analysis. Engineers developing highly automated driving are given the opportunity to achieve a comprehensive, reproducible, and structured test case coverage within a short period of time. AAI ReplicaR’s highly scalable and traceable test methodology is not only beneficial for the development of highly automated driving, but also for the homologation of the respective technology.
Maps & ODDs
Maps & ODDs comes with a representative set of highway and city maps in Germany, while also offering the option to integrate additional maps in OpenDRIVE format. The innovative search function based on the Operational Design Domain is a breakthrough for efficient and focussed testing. Based on the map area chosen in Maps & ODDs, users can proceed with MyScenarios, AutoScenarios or NCAPScenarios for placing the traffic participants and run the simulation.
MyScenarios provides an intuitive interface for the manual creation of scenarios based on the map chosen in Maps & ODDs. ReplicaR users can define basic and advanced parameters, or utilize morphological analysis to efficiently generate and execute scenarios. Key observers and KPIs are used to tag incidents of interest for further analysis of the performance of the system under test (SUT).
AutoScenarios provides an intuitive interface for the automated generation of scenarios based on the map chosen in Maps &ODDs. Placing test vehicles and actors on specific ODDs is done by our unique state-of-the-art synthetic traffic simulation. It is developed using machine learning and generates random and unlimited possible scenarios automatically. Key observers and KPIs are used to tag incidents of interest for further analysisof the performance of the system under test (SUT).
EducAgents includes an integrated reinforcement learning application and thus offers the option to train the traffic agents in AutoScenarios yourself and to optimally adapt them to your tagreted performance. This gives you the freedom and flexibility to create your own intelligent and adaptive systems for the automated creation of test scenarios.
The NCAP Scenarios application comes equipped with the default NCAP scenario catalogue, allowing standardized evaluation of the safety performance of highly automated driving. Special KPIs for analysing the NCAP performance of the system under test are pre-programmed. Leverage the immense potential of simulation to proactively assess and anticipate the NCAP compliance of the system under test (SUT).
Real2Sim allows users to digitize sensor data from real-world test drives for further use in simulation. The process of digitization includes the automatic creation of an OpenDRIVE map of the driven route, including the annotation of objects associated with the map, as well as a log file of the trajectories of detected road users for replay. The application ScenExtract can be subsequently utilized to automatically highlight scenarios of interestand efficiently process them for repeatable testing with MyScenarios or AutoScenarios, ensuring seamless and effective testing procedures.
ScenExtract revolutionizes data analytics with its advanced capabilities. It automatically extracts scenarios based on user-defined requirements from synthetic or digitized test drives. Subsequently, these scenarios can be processed as new test cases with MyScenarios, AutoScenarios and NCAPScenarios.
SenVironment is a powerful application designed for generating detailed and complex static scenes. It features a diverse array of objects, along with material information and comprehensive annotation. The application helps to identify the areas where the accuracy of the sensor perception of camera, LiDar or radar, needs to be improved.
SynthData provides the accurate ground truth data corresponding to the simulated environment created in SenVironment. Our labelled synthetic images and point clouds serve as a valuable reference point for comparing and identifying any variations in the object lists generated by the system under test (SUT) and to generate corresponding KPIs.