OptiCarVis: Improving Automated Vehicle Functionality Visualizations Using Bayesian Optimization to Enhance User Experience

TL;DR

OptiCarVis addresses the challenge of designing AV feedback visualizations that accommodate diverse passengers by combining human input with multi-objective Bayesian optimization. The framework optimizes a 16-parameter AR visualization space to maximize perceived safety, trust, predictability, acceptance, and aesthetics while minimizing cognitive load, using a warm-start or cold-start MOBO strategy and human-in-the-loop feedback. In a between-subject online study (N=117), MOBO-based designs significantly improved perceived safety, trust, and predictability, and reduced cognitive load compared with non-MOBO baselines, though acceptance and aesthetics showed no consistent gains. The work provides empirical evidence that HITL MOBO can efficiently explore large design spaces for in-vehicle UI, enables personalized passenger experiences, and offers an open-source Unity-based implementation for broader adoption and future research.

Abstract

Automated vehicle (AV) acceptance relies on their understanding via feedback. While visualizations aim to enhance user understanding of AV's detection, prediction, and planning functionalities, establishing an optimal design is challenging. Traditional "one-size-fits-all" designs might be unsuitable, stemming from resource-intensive empirical evaluations. This paper introduces OptiCarVis, a set of Human-in-the-Loop (HITL) approaches using Multi-Objective Bayesian Optimization (MOBO) to optimize AV feedback visualizations. We compare conditions using eight expert and user-customized designs for a Warm-Start HITL MOBO. An online study (N=117) demonstrates OptiCarVis's efficacy in significantly improving trust, acceptance, perceived safety, and predictability without increasing cognitive load. OptiCarVis facilitates a comprehensive design space exploration, enhancing in-vehicle interfaces for optimal passenger experiences and broader applicability.

Figures (14)

• Figure 1: Overview of the employed visualizations of an SAE Level 4 SAElevel AV's functional levels of internal operation colley2022scene, CAD mueller2022ar4cad, and status on an AR WSD, showing the possible variations in transparency (alpha) and size values (see brackets). Min and Max represent the designs at the lower and upper bounds of the continuous parameter ranges, while Mid represents the midpoints.
• Figure 2: AV study driving route used in the HITL MOBO iterations (blue) and long route used in the final assessments (orange). Besides, examples of the driver's perspective with all visualizations visible using mid transparency and size values (red).
• Figure 3: The custom parameter design tool allows for adjusting 16 parameters (see \ref{['tab:design_param']}). (1) Users modify values using checkboxes and sliders, with untouched settings highlighted in red. (2) The adjusted values are displayed in a preview that loops the AV driving environment. (3) After interacting with all settings once (their adjustment is optional), (4) the "confirm" button activates. The parameter explanation view (see \ref{['fig:intro']}) is accessible via the "help" button.
• Figure 4: Excerpt of the information given to study participants at the start. Participants were also questioned about the visualizations to ensure understanding.
• Figure 5: Study procedure of the six conditions C1-C6.