E-MailSucheenglishdeutschTwitterYouTubeFacebook

Logo

 

 

The Research Forum groups together various promotion initiatives of the Foundation; these include academic conferences and small-scale research projects. The support is often carried out in cooperation with other foundations and institutions.

 
 

Value-based decision-making 

 

Ethical aspects of decision-making have been at the focus of increased discussion over recent years, as they play an important role in the social acceptance of automated vehicles above and beyond consideration of mere technological aspects. Whilst public discussion has centered above all on situations of dilemma involving unavoidable personal injury, the “Value-based decision-making” (VBDM) project sponsored since 2016 by the Daimler and Benz Foundation considers the influence of ethical issues on everyday decision-making in road traffic. The project is headed by Prof. Dr. Markus Maurer from the Institute of Control Engineering at the Technical University of Braunschweig. To initiate and intensify interdisciplinary dialog between the engineering sciences and the sphere of ethics, the contact to Prof. Jason Millar from the Faculty of Engineering at the University of Ottawa, Canada, was intensified last year.

The research year 2020
Over the past few years, the project’s working group determined values that are relevant for the driving function of an automated vehicle. These relate to the requirements of safety, mobility, comfort, and compliance with road traffic laws and regulations. From this point of departure, attention then moved to the process of accommodating these values in making and generating decisions. Sample scenarios were used for this purpose; one of these is depicted in this illustration of an urban scenario.

This scenario, derived from the Virtual Test Drive (VTD) simulation environment used at the institute, was recorded from the perspective of the automated vehicle. The parked vehicles that are to be passed obscure the view of the roadside, from where a ball out rolls onto the road. The automated vehicle’s scope of action is restricted by an oncoming vehicle. This exemplary scenario illustrates in a nutshell many of the challenges encountered in the development of automated driving functions. One central task relates to dealing with uncertainties in the development process and in all system components during the runtime. The consequences and the resulting demands placed on an automated vehicle were discussed in further detail last year.

In this project year, one focus of research was on the uncertainties within the development process: Our road traffic system is an open world in which practically anything can happen at any time. The specification of a technical system in such application scenarios is therefore a priori incomplete: Even with the most painstaking development, not all safety-critical situations can be defined in advance. Whereas human drivers can fall back on experience and creativity in unexpected situations, this is not immediately possible for an automated vehicle. A potential technical approach to handling uncertain specifications is offered by knowledge-based systems: The aim here is to have the vehicle derive new rules from a set of known rules, and thus at least provide an approach to executing safe behavior in unknown situations.

The situation shown above, in which a ball rolls onto the road from an obscured position, is a well-known example from driving schools. Based on the experience that children play with balls and are likely to carelessly run onto the road after them, a human driver would reduce speed. But without the ability to make knowledge-based decisions, an automated vehicle would have to be provided with fixed rules for precisely this situation. Without such rules, the vehicle could not initiate preventive braking. The aim of a knowledge-based system would be to draw associative conclusions along the following lines: “There is a ball on the road; a child could therefore run out after it.” This association could provide the basis for a decision to brake. In addition to such specific chains of association, however, it is also possible to formulate more abstract factual contexts.

The project participants working together with Prof. Millar and his colleagues from the University of Ottawa have increasingly turned their attention toward meeting this challenge. The Canadian working group carries out research at the interface between the development of technical systems and their configuration under ethical considerations. This cooperation gave rise to a Master thesis that has made an initial contribution to knowledge-based decision-making: The aim was to match requirements from the development process to values, and thus to derive associated parameters and boundary conditions for decision-making during the runtime on the basis of knowledge. This approach is a first step towards a flexible balancing of values in varying situations. The additional knowledge can unite the various elements such as “ball,” “child,” and the resulting accident risk into a causal chain. The threat of an accident is linked to a breach of the value “safety.” To accommodate this value in the resulting vehicle behavior, the boundary conditions for the above scenario can include for example a reduced maximum permissible speed, or a minimal distance to the obscured position. The explicit combination of requirements, values, and boundary conditions thus on the one hand permits an explicit balancing of values, and on the other hand by virtue of the boundary conditions provides an interface to the system of motion planning that had been realized in the previous project year.

Due to the COVID-19 pandemic, implementation of the project findings for demonstration in the experimental vehicle was delayed. Scientific and public discourse also came to a standstill for several months. Further lectures and workshops on this topic were held only at the end of the year, in virtual form. Ethical aspects and possible approaches to dealing with uncertainties were also discussed.

Further planning
In the coming year, the newly generated modules are to be united within the runtime system. Practical tests are expected to yield insights into the mechanism of accommodating values. In addition to being realized in the experimental vehicle, scenarios relevant to this work will be tested and investigated in a simulation environment.

Further information:
Hominin expansions into South East Asia

 

Project management:
Prof. Dr. Markus Maurer, Technical University of Braunschweig, Institute of Control Engineering

Participating scientists:
Susanne Ernst
Marcus Nolte

Website of the project team