A Driving Situation Inference for Autopilot Agent Transparency in Collaborative Driving Context

Overly trust in the autopilot agent has been identi-fied as the primary factor of road incidents involving autonomous cars. As this agent is considered a human driver counterpart in the collaborative driving context, many researchers suggest its transparency to mitigate such overly trust mental model. Hence, this paper aims to develop a driving situation inference method as a transparency provider explaining the types of situations the autopilot agent encounters leading to its certain decision. The proposed method is verified using an autonomous driving simulator called Carla. The findings show that the proposed method can generate situations which enable the human driver to calibrate their trust in the autopilot agent.