Enjoy the Ride Consciously with CAWA: Context-Aware Advisory Warnings for Automated Driving

Erfan Pakdamanian, Erzhen Hu, Shili Sheng, Sarit Kraus, Seongkook Heo, Lu Feng

Abstract

In conditionally automated driving, drivers decoupled from driving while immersed in non-driving-related tasks (NDRTs) could potentially either miss the system-initiated takeover request (TOR) or a sudden TOR may startle them. To better prepare drivers for a safer takeover in an emergency, we propose novel context-aware advisory warnings (CAWA) for automated driving to gently inform drivers. This will help them stay vigilant while engaging in NDRTs. The key innovation is that CAWA adapts warning modalities according to the context of NDRTs. We conducted a user study to investigate the effectiveness of CAWA. The study results show that CAWA has statistically significant effects on safer takeover behavior, improved driver situational awareness, less attention demand, and more positive user feedback, compared with uniformly distributed speech-based warnings across all NDRTs.

Introduction

Level 3 (conditionally) automated driving offers drivers the freedom to disengage from the steering wheel and immerse themselves in Non-Driving Related Tasks (NDRTs), such as reading, gaming, or watching movies. However, this convenience introduces the "out-of-the-loop" problem. When the vehicle reaches its operational limits and issues a Takeover Request (TOR), drivers are often startled and lack the situational awareness necessary to intervene safely.

Traditional warning systems typically rely on uniform auditory beeps. These can be ineffective if the driver is wearing headphones or counterproductive if the sudden noise causes panic. To bridge this gap, we developed CAWA (Context-Aware Advisory Warnings). CAWA provides "gentle nudges"—non-assertive pre-alerts designed to bring the driver back into the loop before a critical emergency occurs by tailoring the warning modality to the driver's specific distraction.

The CAWA Framework

The CAWA system operates through a two-step process of Detection and Selection. First, the system identifies the specific NDRT the driver is engaged in using real-time monitoring. Second, it selects the most effective warning modality based on that sensory context.

To test this framework, we conducted a user study with 20 participants in a high-fidelity CARLA driving simulator. We mapped four common distractions to specific, non-assertive modalities designed to be compatible with the driver's current sensory channel:

  • Mobile Gaming: A text message notification is sent to the phone, as the driver is already looking at the device.
  • Watching a Movie: Vibrotactile (haptic) feedback is delivered through the seat to cut through the visual and auditory immersion of the media.
  • Reading a Physical Book: Speech-based auditory messages are used, as the driver's eyes are directed away from the dashboard.
  • Social Interaction: A Visual HUD (Head-Up Display) projects warnings onto the windshield for a driver who is already scanning the cabin environment.

Experimental Results

The effectiveness of CAWA was compared against a baseline of uniform speech-based warnings across 456 total takeover events. The results demonstrated significant improvements across several safety and performance metrics:

1. Faster and Safer Transitions

CAWA led to significantly faster reaction times across all tested NDRTs ($p < 0.001$). Beyond speed, the quality of the takeover improved; drivers using CAWA exhibited significantly less lateral deviation (lane weaving), indicating more stable vehicle control during the transition.

2. Enhanced Situational Awareness

One of the most striking findings was the impact on "readiness." After receiving a CAWA advisory alert, 87.6% of drivers proactively looked back at the road before the actual TOR was issued. This represents a 26% increase in the likelihood of road monitoring compared to previous studies, effectively mitigating the startle effect of emergency alerts.

3. User Perception and Workload

While the overall cognitive workload remained high due to the inherent difficulty of task-switching, CAWA significantly reduced the "attention demand" subscale. Participants overwhelmingly preferred the system, with 17 out of 20 rating CAWA as "gentler" and "safer" than traditional methods. Some participants did note that haptic feedback could be surprising if unexpected, suggesting a need for further refinement of vibration intensity.

Discussion and Future Work

CAWA demonstrates that the "Modality Match" hypothesis—matching a warning to a driver's current sensory engagement—is a viable path toward safer automated driving. By priming the driver rather than just alerting them, we can reduce the chaos of emergency handovers.

However, limitations remain. This study focused on unimodal warnings; future iterations could explore multi-modal approaches (e.g., combining haptics with visual cues). Additionally, while the fixed 40-second lead time used in this study proved effective, future AI-driven versions of CAWA could dynamically trigger warnings based on real-time risk assessments and the depth of the driver's immersion to prevent alert fatigue.

Conclusion

By transforming warnings from generic alarms into context-aware advisories, CAWA helps drivers "enjoy the ride consciously." This research provides a foundation for more intuitive human-machine interfaces in the transition toward fully autonomous transportation.

Summary updated

Abstract

In conditionally automated driving, drivers decoupled from driving while immersed in non-driving-related tasks (NDRTs) could potentially either miss the system-initiated takeover request (TOR) or a sudden TOR may startle them. To better prepare drivers for a safer takeover in an emergency, we propose novel context-aware advisory warnings (CAWA) for automated driving to gently inform drivers. This will help them stay vigilant while engaging in NDRTs. The key innovation is that CAWA adapts warning modalities according to the context of NDRTs. We conducted a user study to investigate the effectiveness of CAWA. The study results show that CAWA has statistically significant effects on safer takeover behavior, improved driver situational awareness, less attention demand, and more positive user feedback, compared with uniformly distributed speech-based warnings across all NDRTs.
BibTeX
@inproceedings{pakdamanian2022enjoy,
  author = {Erfan Pakdamanian and Erzhen Hu and Shili Sheng and Sarit Kraus and Seongkook Heo and Lu Feng},
  title = {{Enjoy the Ride Consciously with CAWA: Context-Aware Advisory Warnings for Automated Driving}},
  booktitle = {UI},
  year = {2022},
  doi = {10.1145/3543174.3546835},
}
Figure