A Smartwatch-Based Approach to Support and Analysis of Driver Stress and Anxiety

Why driving stress is a problem worth solving

Traffic accidents remain one of the leading causes of death worldwide — roughly 1.19 million fatalities and tens of millions of injuries every year, according to the World Health Organization. In Brazil alone, 2024 saw more than 13,000 deaths and over a million injuries on the roads.

Behind many of these numbers is something less visible than speed or alcohol: the driver's emotional state. Excessive stress degrades driving performance, and driving anxiety is strongly correlated with a poorer quality of life. At the extreme, amaxophobia — the fear of driving — affects an estimated 10% of the world's population, often pushing people to avoid driving altogether, with real consequences for their mobility and independence.

This is especially acute for new drivers. Learning to drive is stressful by nature, and that stress feeds back into mistakes, which feed back into more stress. If we could help student drivers recognise and manage their anxiety while they drive, we might make them safer — and help them earn their licence faster.

The idea: let the watch do everything

Most physiological monitoring research leans on signals like pupil dilation, galvanic skin response, or blood pressure. These are powerful, but they typically require specialised hardware, expert setup, and often a second device (a phone) to synchronise with. That cost and complexity puts them out of reach for everyday use.

We took a deliberately pragmatic stance: heart rate is the most widely studied stress signal, and modern smartwatches already measure it well. So we built the entire system to run on a smartwatch alone — no phone pairing required.

A smartwatch that runs Android (Wear OS) apps, reads heart rate, has GPS, and connects to the internet over its own eSIM gives us everything we need to monitor a driver unobtrusively, in real time, on the road.

What the project contributes

• A Wear OS application that monitors and supports drivers by collecting heart rate and GPS coordinates during driving, storing the data in the cloud in real time, and giving instant feedback to help manage stress and anxiety.
• A reporting tool that summarises each driver's heart rate and route so specialists — such as psychologists — can review the data and offer individualised care.
• Open source code, published so other researchers and developers can review and extend the system: github.com/motaoliveiraufpr/CFCStress.
• A pilot case study with eleven student drivers, conducted during real practical lessons under procedures approved by an ethics committee.

How the system reads stress

There is no universal standard for mapping heart rate onto "stress." Comparing a participant's heart rate against their own baseline is a recognised, practical approach in the literature, so we defined five zones empirically, based on beats per minute (BPM):

These thresholds are a starting point, not a verdict — a clear avenue for future work is to calibrate them per driver using the BPM actually collected in experiments.

The app, from the driver's seat

The application was built for Wear OS using the Samsung Health Sensor API (for compatibility), written in Kotlin and Java, with Firebase handling real-time storage and synchronisation. Audio feedback was produced with a text-to-speech tool, voiced by a friendly young female persona named Juliana — a deliberate choice to make the interaction feel warmer and more personal.

A driving session flows like this:

1. The driver creates an account and logs in on the watch.
2. On start, the app greets them by voice — "Welcome, my name is Juliana, I'm your virtual assistant" — to humanise the experience.
3. Every three seconds, it records heart rate and GPS coordinates, saving them asynchronously to the cloud, and displays the current heart rate and stress zone on screen.
4. Every six minutes, it plays an audio message matched to the driver's current zone.
5. In the high zones (Z4 or Z5), a prompt also appears on screen offering a guided breathing exercise — if the driver has safely stopped the car and accepts, the relaxation audio plays.
6. When the lesson ends, the driver presses stop, which closes the session and finalises the stored data.

The feedback, in its own words

The feedback escalates gently with the stress zone — reassurance when calm, a nudge toward deep breathing when anxiety begins, and a clear instruction to pull over when anxiety is high:

Once the driver has safely stopped, the relaxation routine walks them through diaphragmatic breathing — inhaling for five counts, holding, and exhaling slowly — repeated for a couple of minutes. Breathing techniques like these are well established in the clinical literature for managing acute stress and anxiety.

Turning a drive into a map specialists can read

Real-time coaching helps the driver in the moment. But the second half of the contribution is what happens afterwards.

Using a Python pipeline built on GeoPandas and Folium, the system connects to the cloud database and plots the entire route on an interactive Google Maps layer. Each GPS point is coloured by the stress zone the driver was in at that location:

The result is a vivid emotional map of the lesson: a specialist can see exactly where on a route a learner's anxiety spiked — a particular junction, a busy avenue, a tricky manoeuvre. Because reports are generated while the driver is still on the road and can be accessed from anywhere with internet, a professional can even intervene immediately after (or, in extreme cases, during) a session.

This two-layer architecture — a mobile layer for real-time feedback and self-regulation, and a desktop layer for longitudinal, multi-user supervision by professionals — mirrors the monitoring → interpretation → intervention pattern recommended in multi-device human–computer interaction research.

The pilot study

To validate the tools, we ran a pilot case study with eleven student drivers from a Driver Training Centre (the Auto Escola Milênio), during their real practical lessons. We deliberately chose novice drivers because they are precisely the population most likely to experience high stress and anxiety, and because helping them could shorten the path to a licence.

The procedure, approved in advance by an ethics committee for research with humans:

1. All students taking practical lessons were invited; instructors helped identify those most likely to show signs of stress.
2. Each selected student signed a free and informed consent form.
3. Participants wore the watch on their wrist, logged in, and completed a normal lesson — about 40 minutes of driving — accompanied by an instructor.
4. Afterwards, students and instructors answered a questionnaire, and a psychologist specialising in fear of driving evaluated the reports produced from the collected data.

Before driving, students were also shown a short mindfulness primer — present-moment awareness, observing the environment, focusing on the breath, and accepting thoughts without judgment — plus a guide to diaphragmatic breathing. The intent was to make the in-the-moment feedback land better: a driver already familiar with the technique is more receptive when the watch asks them to use it.

What we found

All eleven participants completed every activity and reported no problems wearing the watch. The questionnaire results were encouraging:

Notably, none of the students reached zones Z4 or Z5 during the pilot — yet they still reported feeling less anxious when using the app, saying the feedback made them feel more comfortable and confident behind the wheel. Two of the eleven were interested enough to want to buy the watch and keep using the application beyond the study.

The six instructors who supervised the lessons observed a slight increase in students' confidence and a positive effect on the learning experience. The specialist psychologist underlined that confidence is an essential factor for learning to drive — and that the maps produced are a genuinely valuable tool for analysis. Both instructors and the psychologist suggested refining the feedback (its text was sometimes triggered too frequently) and re-evaluating the stress zones against the collected data.

Conclusions and where this goes next

This study introduced a smartwatch-only solution for monitoring and analysing driver stress and anxiety from heart rate, with no second device required. It delivers real-time, self-regulating feedback to the driver, and stores everything in the cloud so specialists can review each session and provide personalised support.

The pilot demonstrated feasibility and a real potential to improve driver well-being — particularly for novices. The limitations are honest ones: a small sample, and stress thresholds that need per-driver calibration. Future work points clearly toward:

• Larger and more diverse studies, including experienced drivers who report high stress.
• Interactively tuning the stress zones from real collected data.
• Adding other low-cost wearable signals (galvanic skin response, respiration rate) to improve accuracy.
• Applying machine learning to predict when a driver is about to enter a high-stress zone — and to deliver feedback before it happens.

Acknowledgments

This work was supported by the Coordination for the Improvement of Higher Education Personnel (CAPES) — Program of Academic Excellence (PROEX). We thank the Driver Training Centre Auto Escola Milênio, where the experiments were conducted, and psychologist Juliana Daga for her continuous involvement throughout the project.

How to cite

de Oliveira, T. M., Ortoncelli, A. R., Casa, C., Casa, C. and Silva, L. (2025). A Smartwatch-Based Approach to Support and Analysis of Driver Stress and Anxiety. In Proceedings of the 21st International Conference on Web Information Systems and Technologies (WEBIST 2025), pages 544–552. DOI: 10.5220/0013717000003985.

This is a plain-language adaptation of the peer-reviewed paper, prepared for a general audience. For the full methodology, figures, and complete reference list, please consult the original publication.