Trust-Aware Control of Automated Vehicles in Car-Following Interactions with Human Drivers

TL;DR

This paper develops a quantitative trust dynamic model for human drivers interacting with automated vehicles, integrating trust into AV control to improve plan explicability and increase human trust in traffic scenarios.

Contribution

It introduces a novel trust dynamic model based on explicability and incorporates it into AV control design for better human-AV interaction.

Findings

Trust-aware AVs generate more explicable plans.

Trust-aware AVs achieve higher human trust levels.

Trust modeling improves AV decision-making in traffic.

Abstract

Trust is essential for automated vehicles (AVs) to promote and sustain technology acceptance in human-dominated traffic scenarios. However, computational trust dynamic models describing the interactive relationship between the AVs and surrounding human drivers in traffic rarely exist. This paper aims to fill this gap by developing a quantitative trust dynamic model of the human driver in the car-following interaction with the AV and incorporating the proposed trust dynamic model into the AV's control design. The human driver's trust level is modeled as a plan evaluation metric that measures the explicability of the AV's plan from the human driver's perspective, and the explicability score of the AV's plan is integrated into the AV's decision-making process. With the proposed approach, trust-aware AVs generate explicable plans by optimizing both predefined plans and explicability of the plans in the car-following interactions with the following human driver. The results collectively demonstrate that the trust-aware AV can generate more explicable plans and achieve a higher trust level for the human driver compared to trust-unaware AV in human-AV interactions.