Currently, various possible solutions are emerging to replace the perception performance of train drivers (Tf) with technical solutions for autonomous driving operation. However, in the higher levels of automation (GoA) 3 and 4, it is envisaged that the responsibility for safety no longer lies with a TF. Various approaches to replace the perceptual performance of the Tf are not suitable according to the current state of affairs. For example, implementation with the help of ETCS Level 3 plus fencing, the transformation of the entire railway network into a fully automated subway network, so to speak, is not an alternative due to a variety of factors. Likewise, trackside sensors, predictive drones or chip-based tracking of all potential obstacles are not yet ready for the market. The further development of on-board contactless perception systems as a possible solution has been dormant for about 20 years at technology maturity level (TRL) 5. Contact-based perception systems such as rear-end collision and impact detection are also unsuitable for railways due to their systems.

Multisensory perception systems have emerged as a promising approach at present, but they must be evaluated in terms of suitability and eligibility. In order to approach the problem in a structured way, the DZSF project "Risk Acceptance Criteria for Automated Driving on the Rails" dealt with the definition of risk acceptance criteria for the automation functions that will take over these tasks of the Tf in the future. Essentially, this resulted in a function list with associated boundary conditions and safety requirements. As a result, safety requirements in the range from SIL 1 to SIL 2 were derived for obstacle detection tasks. Based on these results, the question arises as to how these derived requirements can be compared with the conditions and hazards of a specific route.

One challenge is the lack of appropriate performance metrics and understandable performance requirements that can be used to evaluate perception systems and their performance through derived security requirements. These requirements depend largely on the defined area of application or the Operational Design Domain - (ODD for short).

In this project, a concept for the comparability of performance metrics for perception systems with ODD-specific characteristics is to be developed. Finally, on the basis of researched security requirements, concrete performance requirements are to be derived or estimated.

Please refer to the published documents for the details of the tender, in particular the service description and the bidder information.