Effective verification of automated driving – Daimler presents the PEGASUS research sub-project "Testing"
- Development of uniform standards for the verification of highly automated vehicle systems
- Establishing a testing method as a worldwide standard
- Development of a model for real traffic situations that fully describes all scenarios
Stuttgart/Wolfsburg. PEGASUS (Project for establishing generally acccepted good quality criteria, tools, methods, scenarios and situations for the approval of highly automated driving functions) is a research project subsidised by the Federal Ministry of Trade and Energy (BMWi). The aim of the joint project is to develop uniform technical standards for the verification of highly automated vehicle systems, and to answer important questions concerning the safety and reliability of the systems. On 13 and 14 May, to conclude the research project,Daimler AG together with 16 other project partners is for the first time presenting the most important findings from the project to the German and international public in Wolfsburg.
The dream of the self-driving car has perhaps never been as tangible as now. In purely technical terms, all the conditions for it are in place. However, many questions still need to be resolved before automation systems can be used on a general basis and in millions on the roads. Among them are how the safety and reliability of these systems can be verified, when a system is "mature" enough for general approval, or what role humans will play in interaction with the technology. "Accordingly we played a major part in PEGASUS to help develop a state-of-the-art methodology that is accepted by all interested parties worldwide. This test methodology will help us to bring reliable and robust, highly automated vehicle systems onto the road for our customers," is how Michael Hafner, Head of Driving Technologies and Automated Driving at Mercedes-Benz Cars Development, explains the importance of PEGASUS for the secure future of autonomous driving.
Daimler is responsible for the central PEGASUS research sub-project "Testing"
PEGASUS fundamentally consists of four sub-projects, each of which covers an important topic area in the verification of autonomous driving and resolves the relevant questions:
- TP 1 – Scenario analysis and quality standards
- TP 2 – Implementation processes
- TP 3 – Testing
- TP 4 – Reflection on results and embedding
Sub-projects 1 and 2 are primarily concerned with the definition of relevant traffic scenarios and the critical analysis of already existing verification processes for automated driving systems in the automobile industry. With around two thirds of the total project volume, sub-project 3 "Testing" is the largest sub-project in PEGASUS. Under the aegis of Daimler AG, its aim is to develop methods and tools for conducting tests by simulation, on test sites and in real road traffic.
"To appreciate the relevance of this sub-project, it is important to understand the previous role and current status of test procedures in the automobile industry. To date onboard driving assistance systems and active safety systems are tested over hundreds of thousands of kilometres in real traffic operation and in several hundred different traffic scenarios on test sites. If they show no relevant functioning errors, they can be approved. However, to establish a comparable level for highly automated driving systems, we would require billions of test kilometres," says Dr Helmut Schittenhelm, PEGASUS project manager for "Testing" at Mercedes-Benz Cars Development, and adds: "Only simulations with specifically chosen parameters can bring this down to an acceptable and realistic level of test kilometres on the roads. This means that in the future, simulations will take on a central role in test procedures and trials. Developing and establishing a reliable methodology for this is one of our main tasks in PEGASUS."
PEGASUS was called into being in January 2016, following an initiative by the Federal Ministry of Trade and Energy (BMWi). A total of 17 project partners, among them many German vehicle manufacturers, suppliers, public institutions and technical colleges and faculties have since then devoted themselves to four sub-projects using a function called "Autobahn Chauffeur" as an application example.
One key element of the sub-project "Testing" was the development of a model for real traffic situations that fully describes all variations and influencing factors. These scenarios and the test case database were put together using the "Autobahn Chauffeur" system as an example, in close cooperation with the research institute "Forschungsgesellschaft Kraftwahrwesen mbH Aachen (FKA)" and the "Institute for Motor Vehicles at the technical university in Aachen (IKA). To this end, challenging situations were collated in this task package for PEGASUS, with systematic and highly precise measurement of the relevant parameters such as speed, directional stability or route conditions. In future test procedures, this scenario database is intended to ensure and verify that in the present state of the art, the Autobahn Chauffeur responds reliably and robustly in the interests of preventing collisions. The aim of the database is internationalisation in the sense of interfaces and a modelling language for scenarios, so that these can be collated, exchanged and compared.
Subsequent simulation tests were the second task package in the sub-project "Testing". The aim here was to develop solutions for systematic testing of the system limits within the set parameters, so as to ensure that an autonomous system also responds as the limits are approached. This requires intelligent networking of the simulations with experiments on test sites and in field tests. The aim of the simulation is to identify collision-related scenario variations (e.g. remaining distance) according to parameter variations (variety of scenarios), and then repeatedly verify the simulation results with specific tests on the test site.
In the third task package "Test methodology", a testing technique was developed eventually allows scenarios to be precisely reenacted on the test site with programmable test objects. Sebastian Werr, Verification of Test Technology and Communication Assistance Systems, Mercedes-Benz Cars Development, on the test methods employed in these tests: "For the self-driving targets, the robot vehicles, we apply a future-oriented control technology which Daimler already developed with external partners around seven years ago. The aim of this is to realise precisely prescribed test scenarios. Not only can the chosen scenarios be reproduced at will, they can also be reproduced with absolute accuracy. This means that the system vehicle can absolve identical tests as often as required – even weeks later if the developers have made changes."
Testing the scenarios in real field trials rounds the PEGASUS sub-project "Testing" off as the fourth task package. Field tests are a useful addition in cases where the models are not yet comprehensively mature. A live presentation of the control technology will be one of the main items in the programme for the final PEGASUS event in Wolfsburg, and underpin the technological advances and international importance of the joint PEGASUS project.
During the course of the almost three-year project, PEGASUS has become an international brand – the definitive term for establishing a test methodology to verify highly automated vehicle systems. PEGASUS is symbolic of the development of test procedures to be absolutely sure that automated driving systems can be reliably used all over the world. This is another reason why since the onset of the project, there has been a regular exchange of information with numerous automobile manufacturers, government bodies and associations in Europe, the USA, Japan and China.
Creating confidence. Ensuring reliability.
There is an ongoing public discussion on the subject of "autonomous driving", especially at the ethical level. Accordingly Daimler AG is particularly keen to create confidence in the new technologies and overcome public reticence. "It is relatively simple to install a sensor and computer in a car that keeps it on track in favourable environmental conditions. But also to ensure this systematically when other factors are in play, e.g. a low sun angle or in wind and rain, is very challenging and extremely important for dependable use of a vehicle. Such verification has always been standard in the automobile industry. But it is precisely this public discussion that casts the spotlight on the USP of car manufacturers," says Dr Schittenhelm. "With PEGASUS we are not only making a technological contribution to the verification of autonomous driving,but also creating public confidence in the reliability of the technologies employed."
In July 2019, even before this project officially ends, several follow-up projects likewise approved and subsidised by the BMWi will start. These will mainly involve transferring the findings from the topic context of "Autobahn" to the development of verified test procedures for relevant scenarios in urban traffic. From 1 July one of the projects in which Daimler AG is heavily involved is a project with the previous working title "VV Methods" (validation and verification methods). As PEGASUS is to be retained as a brand, it will probably continue under the project name PEGASUS 2.0.
More information about the PEGASUS project can be found at www.pegasusprojekt.de/de/.