Safety and ethical concerns in mixed human-robot control of vehicles

Abstract

Many robotic applications require human-like behaviour of an artificial agent and quite often include mixed (human-robot) control set-up for effective operation and to meet regulatory requirements. Developing such systems with high level of autonomy implies a good understanding and variability of human behaviour. One of the most popular areas of research in robotics with mixed control is to develop self-driving cars that are able to participate in normal traffic scenarios and acceptable under established risk management processes. In Berlin, some cars have already been licensed, but manufacturing driverless cars is more difficult than usually assumed. This is mostly because vehicular control should be human-like to avoid confusing pedestrians, passengers or other human drivers. Among the many difficulties to achieve human-like control in ensuring safety requirements are satisfied as well as being ethically acceptable, the problem of identifying and calibrating control parameters is far more complex than traditional control and systems theory alone would be able to handle. The paper provides insights into the difficulties of autonomous and mixed vehicle control and generally warns about the theoretical and ethical consequences of our limited understanding of human performance issues in car driving.