Trust-aware Safe Control for Autonomous Navigation: Estimation of System-to-human Trust for Trust-adaptive Control Barrier Functions

TL;DR

A trust-aware safe control system for autonomous navigation in the presence of humans, specifically pedestrians, that combines model predictive control with control barrier functions (CBFs) and system-to-human trust (SHT) estimation to ensure safe and reliable navigation in human-populated environments is presented.

Abstract

A trust-aware safe control system for autonomous navigation in the presence of humans, specifically pedestrians, is presented. The system combines model predictive control (MPC) with control barrier functions (CBFs) and trust estimation to ensure safe and reliable navigation in complex environments. Pedestrian trust values are computed based on features, extracted from camera sensor images, such as mutual eye contact and smartphone usage. These trust values are integrated into the MPC controller's CBF constraints, allowing the autonomous vehicle to make informed decisions considering pedestrian behavior. Simulations conducted in the CARLA driving simulator demonstrate the feasibility and effectiveness of the proposed system, showcasing more conservative behaviour around inattentive pedestrians and vice versa. The results highlight the practicality of the system in real-world applications, providing a promising approach to enhance the safety and reliability of autonomous navigation systems, especially self-driving vehicles.

Figures (16)

• Figure 1: Example images of pedestrians engaged with their smartphones in a variety of poses, backgrounds, and lighting conditions
• Figure 2: Model for Smartphone Usage Classification - EfficientNet V2 feature extractor followed by a three-layered fully-connected classification head
• Figure 3: Eye contact detection using pose estimation belkada2021pedestrians - green pose skeleton on the left image indicates presence of eye contact while the red pose skeleton on the right image indicates otherwise
• Figure 4: The complete trust estimation from captured RGB images of pedestrians - Memory in analogous to history i.e. the estimator stores the trait scores, trust values, and aggregated trust for at least one timestep in the past
• Figure 5: Comparison of different trust dynamics - with raw estimates of trust based on confidence scores only