How to enable safe testing of automated vehicles at proving grounds - answers to be found in the new SAFER associated project ETAVEP
Enablers for Testing Autonomous Vehicles at Existing Proving ground – ETAVEP, is the name of a new associated project in SAFER’s project portfolio. At today’s proving grounds, safety during testing is guaranteed by trained skilled test drivers. However, future autonomous vehicles will either lack controls or simply not have room for a skilled test driver. Thus, it is now essential to develop effective methods to ensure that testing is safe for the people who work on the test track.
The project aims to research which technologies and which level of technology performance is required to replace the safety driver with another solution for autonomous vehicle testing at existing proving ground facilities while maintaining the manual driven vehicles, mixed traffic. The project started in April and will be concluded in March 2022. Volvo Cars, AB Volvo, AstaZero, RISE, Chalmers, SafeRadar are partners in the project. Albert Lawenius, at Volvo Cars is leading the project.
A summary of the research questions that will be addressed in the project:
1. Which global monitoring principles need to be applied? To what extent does the current proving ground traffic control concept need to be extended and/or changed to adequately supervise autonomous vehicles on test tracks?
2. Which local monitoring principles need to be applied? Since both living and inanimate objects may appear on the test track unexpectedly (in the sense of a tree falling down, an animal jumping the fence or a road worker experiencing transponder malfunctions), which object detection and classification capabilities need to be installed in addition to those used for continuous traffic control?
3. Can a type of on-board monitoring for vehicle faults be developed that detects mechanical faults and wear as well as (or better than) an experienced test driver? In particular, will statistical models of expanded non-parametric transmissibility estimates using Local Rational Models (LRM) with a general sensor setup provide a sufficiently robust and accurate performance in this capacity?
4. How to take emergency control over an autonomous vehicle at risk? If a risk materialises in any of the above monitoring systems, how does one make sure that the autonomous vehicle can be brought to a safe stop?
5. What is a sufficient set of test cases for validating concepts that have been developed to address 1-4? Since it is impossible to test for all combinations of potential errors, what is an appropriate set of edge cases that if handled will guarantee the desired safety envelope in the four safety aspects described above?