Ethics Preference Modeling and Implementation of Personal Ethics Setting for Autonomous Vehicles in Dilemmas

Autonomous vehicles (AVs) inevitably confront dilemmas in which collisions cannot be avoided, and collision avoidance involves ethical issues due to harm distribution among traffic participants. Regarding how AVs should respond to dilemmas, ethical divergence is the main barrier that exists universally in human society. To overcome this barrier, the rational ethical propensity, i.e., the inclination of evasive steering (IES), of Chinese respondents in two typical dilemmas was collected via an online survey to construct AVs' ethical model. Respondents were partitioned into two subgroups with relative ethical preferences to address the main ethical divergence based on their IES. The analytical IES ethical model, describing relationships among the respondents' ethical intention and sensitive factors, was proposed. Three IES models for the two partitioned subgroups and the total respondents were fitted. Personal ethics setting (PES) was implemented with the subgroups' models as the boundaries of operation space. PES-based trajectory planning experiments were conducted, and simulation results showed that AVs make evasive decisions in line with their ethical preference. © 2009-2012 IEEE.