We are excited to share that work done in the VRLab (CISUC/DEI) will be presented in the 18th ACM International Conference on Automotive User Interfaces and Interactive Vehicular Applications (AutomotiveUI Adjunct ’26), taking place this September in Gothenburg, Sweden.
As part of our work under the ADSafeVANET research project, our team explored how human drivers in legacy vehicles interact with automated vehicles through Vehicle-to-Vehicle (V2V) cooperation requests.
Evaluating Multimodal Touchscreen Interfaces for V2V Overtaking Requests in Virtual Reality
Authors: Rodrigo Rodrigues and Jorge C. S. Cardoso
As autonomous vehicles (AVs) and human drivers share the road, establishing clear communication during complex maneuvers — such as cooperative overtaking — is crucial for safety. In this study, we investigated how drivers process and respond to incoming V2V overtaking requests displayed on a center-stack touchscreen interface within an immersive Virtual Reality driving simulator.
To simulate real-world divided attention, 26 participants performed a primary driving and visual distraction task while receiving two types of requests: normal overtaking requests and time-sensitive requests requiring quick decisions. We evaluated three distinct alert modalities across the touchscreen UI:
- Visual-only (V): Standard on-screen popups.
- Visual + Haptic (VH): Screen popups paired with vibrotactile seat cues.
- Visual + Auditory + Haptic (VAH): Full multimodal alert stacking using screen popups, audio chimes, and backrest vibrations.
Key Findings & Human Factors Insights
Our empirical evaluation revealed a compelling human-machine interaction trade-off:
- Speed & Accuracy Gains: The full multimodal configuration ($VAH$) significantly improved driver response times during normal requests and dramatically increased request-type classification accuracy for time-sensitive requests (rising to 57.7% accuracy compared to just 32.3% in visual-only).
- The Psychological Cost: Stacking sensory modalities created a cascading escalation in perceived situational pressure, temporal demand, and driver frustration during time-sensitive requests, along with a marginal increase in perceived system intrusiveness.
These findings highlight a central design paradox for cooperative driving interfaces: while multi-sensory density recovers tactical driver accuracy, it imposes an immediate cognitive and psychological load. This research points toward future HMI architectures that leverage synchronized, low-latency multi-sensory alerts to preserve decision accuracy while minimizing driver stress.
We look forward to connecting with automotive HCI researchers and mobility pioneers in Gothenburg this September!