One of the oft repeated ethical dilemmas associated with autonomous vehicles relates to scenarios where a crash could occur. Should the car veer towards a crowd in an action that might save the driver but injure other people, or should the car turn towards a wall potentially killing the driver but saving others from harm?
Should the guiding principle be to protect as many people as possible at the expense of fewer people? But what if the car is carrying someone of importance, like a president?
A new algorithm cuts the middle path by seeking to distribute risk ‘fairly’ (and by ‘fairly’ this means equally). The German research utilises software that makes more ethically differentiated decisions.
The Technical University of Munich (TUM) study has an algorithm contained in software which incorporates the 20 ethics recommendations of the EU Commission expert group for autonomous driving. This is by assessing the varying degrees of risk to pedestrians and motorists and making a decision. The researchers have made the code available to the general public as Open Source software.
To test out the software, the researchers ran through some 2,000 scenarios involving critical situations. These were distributed across various types of streets and regions such as Europe, the U.S. and China.
Discussing the study, lead scientist Maximilian Geisslinger, who is the TUM Chair of Automotive Technology, says: “Until now, autonomous vehicles were always faced with an either/or choice when encountering an ethical decision. But street traffic can’t necessarily be divided into clear-cut, black and white situations; much more, the countless grey shades in between have to be considered as well. Our algorithm weighs various risks and makes an ethical choice from among thousands of possible behaviours — and does so in a matter of only a fraction of a second.”
The ethical basis to the algorithm is to give priority for the worst-off and a fair distribution of risk among all road users. Various factors influence this: A truck, for example, can cause serious damage to other traffic participants, while in many scenarios the truck itself will only experience minor damage. Yet the opposite is the case for a bicycle.
The researchers also added variables to the calculation which account for responsibility on the part of the traffic participants, for example the responsibility to obey traffic regulations.
One of the scenarios is as follows:
An autonomous vehicle wants to overtake a bicycle, while a truck is approaching in the oncoming lane. All the existing data on the surroundings and the individual participants are now utilized. Can the bicycle be overtaken without driving in the oncoming traffic lane and at the same time maintaining a safe distance to the bicycle? What is the risk posed to each respective vehicle, and what risk do these vehicles constitute to the autonomous vehicle itself?
In unclear cases the autonomous vehicle with the new software always waits until the risk to all participants is acceptable. Aggressive manoeuvres are avoided, while at the same time the autonomous vehicle doesn’t simply freeze up and abruptly jam on the brakes.
The options of ‘Yes’ or ‘No’ are considered to be irrelevant and instead they are replaced by a far more sophisticated response, based on evaluations containing a large number of options. In the future the algorithm will need to be strengthened in order to consider the types of differentiations the exists between different societies and cultures.
The research appears in the journal Nature Machine Intelligence, titled “An ethical trajectory planning algorithm for autonomous vehicles.”
