Why this paper is useful

Although this paper is mainly about cognitive therapy and anxiety rather than autism or airports, it has a lot of crossover with my project. It introduces Virtual Reality Exposure Therapy (VRET) and explores how VR can be combined with physiological data such as heart rate, HRV, EEG and EDA to create personalised, adaptive experiences. It also briefly mentions applications involving autism spectrum disorders.

The most useful idea is that VR can respond to the person rather than remaining static. Physiological signals can be monitored in real time and used to change elements of the virtual environment depending on the user's emotional state.

Why VR could work well for airports

The paper describes VR as “safe, highly controllable, repeatable, and private.” These qualities are particularly relevant to airport environments, which can be unpredictable, crowded and difficult to recreate repeatedly for testing. VR could allow noise, crowds, lighting or other sensory factors to be introduced gradually and altered depending on the individual.

Another useful argument is that VR does not necessarily need to be better than real-world exposure. One study describes it instead as a more practical and versatile method where real situations are difficult or expensive to recreate. This could help justify using simulated airport environments.

Personalisation and biofeedback

A major connection to my project is the use of biofeedback. Heart rate, HRV, EEG, EDA and breathing can be monitored instead of relying only on somebody verbally explaining how anxious or overwhelmed they feel. The paper describes this as moving beyond subjective self-report and creating a dynamic feedback loop between the user's physiological state and the VR environment.

This could be particularly interesting for my autism research, but I would need a separate autism-specific source before arguing that autistic adults have particular difficulty identifying or communicating these internal states.

A useful example in the paper is Agora-VR, where crowd noise and illumination are automatically changed according to heart-rate feedback. This feels particularly relevant to an airport-based sensory environment.

Visualisation

Traditional imagined exposure relies on somebody being able to visualise a feared situation. The paper notes that not everybody can do this effectively, whereas VR can create a multisensory experience without relying completely on imagination.

This might connect to autism, but again I need an autism-specific source before claiming that difficulty visualising is particularly common among autistic adults.

Important distinction for my project

Traditional VRET normally works by putting somebody into a controlled simulation, deliberately exposing them to anxiety-provoking stimuli and helping them gradually regulate their response.

My idea may be different.

I am potentially looking at somebody who is already inside a real airport and beginning to become overwhelmed, with technology detecting this change and then helping them regulate.

So my project might eventually be better described as adaptive VR, biofeedback-responsive VR, or an immersive sensory regulation intervention, rather than VRET. This is something I need to establish through further research.

VR, AR or MR?

The paper also discusses VR, augmented reality and mixed reality as different immersive approaches.

This makes me question whether full VR is actually the most appropriate solution inside an airport. If someone still needs to walk, navigate and remain aware of their surroundings, AR or MR might ultimately make more sense. I should keep this open rather than deciding on VR too early.

Alpha and Beta brain activity

One referenced study found that decreased alpha and increased beta activity were associated with higher cognitive control in VR.

This is interesting and worth researching further, but the paper does not support a simple idea that alpha = calm and beta = anxiety. Therefore, I should not put “shifting between Alpha and Beta” into my research question yet. EEG could eventually become one possible method of measurement, but more evidence is needed first.

Research gaps and limitations

The paper identifies several weaknesses in existing research: small and non-diverse samples, limited long-term research, lack of standardised evaluation, reliance on limited physiological signals and the technical difficulty of combining VR, AI and biofeedback. It also identifies equipment cost, accessibility and cybersickness as barriers.

For my project this creates additional questions around whether a VR headset itself could cause sensory discomfort, whether it is safe or practical inside an airport, and whether putting someone already overwhelmed into another immersive environment could actually make things worse.

How the system works

The paper's proposed system follows a useful basic loop:

physiological data → process data → identify emotional state → change VR environment → measure response → repeat.

For example, it could classify somebody as Calm / Mild Anxiety / Escalated and then reduce crowd density or alter lighting accordingly.

This also shows that a fully automatic AI system is technically complicated. For my master's, I probably do not need to build this entire system. A simplified prototype using something like heart-rate threshold → environmental change could demonstrate the concept without turning the project into a machine-learning project.

What this changes in my project

The strongest emerging direction is no longer simply: How can design make airports less overwhelming?

It is moving towards: How might adaptive immersive technologies and physiological biofeedback support autistic adults in regulating anxiety and sensory overload during airport travel?

I would keep this wider wording for now rather than locking myself into VRET, EEG or Alpha/Beta.

Key Terminology

There was some really great terminology in this paper that could be useful for me.

Virtual Reality Exposure Therapy (VRET) - Exposure therapy delivered through immersive VR environments. It allows potentially anxiety-inducing situations to be simulated in a controlled setting rather than experienced entirely in real life.

Biofeedback - Using information collected from the body to understand and potentially influence the user's state. The paper uses measures such as heart rate, HRV, EEG and EDA.

Physiological Monitoring - Measuring bodily signals in real time to provide information about emotional or cognitive responses rather than relying entirely on self-report.

Heart Rate Variability (HRV) - Variation in the timing between heartbeats. It is one of the physiological signals discussed for monitoring and responding to changes in the user's state.

Electroencephalography (EEG) - Measurement of electrical activity in the brain. The paper discusses EEG alongside other physiological measures and later refers to changes in alpha and beta activity.

Electrodermal Activity (EDA) - Measurement associated with changes in skin electrical activity. Used alongside other signals as a way of identifying physiological responses.

Galvanic Skin Response (GSR) - Another skin-response measure used in one of the adaptive VRET systems described in the paper alongside heart rate.

Multimodal Data / Multimodal Sensing - Combining several forms of information rather than depending on one measurement. Examples include HRV, EEG, EDA, respiration, voice and movement. The paper argues this can provide a richer understanding of the user's state.

Closed-loop System / Adaptive Feedback Loop - A continuous cycle where the user's physiological state is measured, the system responds, their new reaction is measured, and the environment changes again. Very important terminology for my project.

Real-time Adaptation - Changing the VR environment immediately in response to information being collected from the user. For example, reducing environmental intensity if anxiety becomes high.

Patient-centred Adaptive VR Framework - The paper's terminology for an adaptive VR system centred around the individual's physiological response. For my project I would probably change patient-centred to user-centred, because autistic travellers are not necessarily patients.

Cognitive Control - The ability to regulate behaviour and cognitive processes in response to demands. One referenced study examines how emotional arousal and physiological activity relate to cognitive control inside VR.

User Autonomy / Self-regulation - The paper discusses adaptive biofeedback as a way of supporting the individual in regulating their own emotional or cognitive state. This could be important to my project's focus on independent airport travel.

Alpha / Beta Oscillatory Activity - Different patterns of EEG activity. The paper reports one study linking decreased alpha and increased beta activity with higher cognitive control, but this should not yet be interpreted simply as “alpha = calm / beta = anxious.”

Large Language Model (LLM) - A form of AI mentioned in the paper in relation to personalised therapeutic systems and generating information from collected data.

Machine Learning (ML) - Algorithms that identify patterns within physiological or behavioural data, including patterns associated with anxiety or stress.

Reinforcement Learning - A system that learns which environmental changes work best based on whether the user's physiological response improves or worsens.

References from this paper worth following

12] Y. She, F. Liu, B. Yang, and B. Hu. Converging Real and Virtual: Embodied Intelligence-Driven Immersive VR Biofeedback for Brain Health Modulation.

24] M. Li, J. Pan, Y. Li, Y. Gao, H. Qin, and Y. Shen, Multimodal physiological analysis of impact of emotion on cognitive control in VR.

[17] M.-H. Tayarani-N. and S. I. Shahid Detecting anxiety via machine learning algorithms: A literature review.

[14] S. Ge, S. Luo, S. Yan, and X. Shen, Effects of Augmenting Real-Time Biofeedback in An Immersive VR Performance.