Development of a Virtual Reality Flight Simulation and Control Interface for Adaptive Exposure Therapy using the BRAVE Framework

BRAVE (Bioadaptive Regulation through AI-based Virtual Exposure) is a multidisciplinary research project that aims to improve Virtual Reality systems used in the early treatment of Specific Phobias (SPs) by combining computational modelling, neuroscience and clinical psychology. The initiative seeks to develop an adaptive therapeutic framework that integrates neurofeedback and automatic emotion recognition to personalise exposure-based interventions. BRAVE is conceived as a closed-loop therapeutic system in which patients with phobias are exposed to safe and controlled virtual environments while their biometric feedback, including physiological and ocular responses such as heart rate and pupil size, is continuously recorded by a multimodal sensing system. These biosignals are transmitted either to a human therapist or to an AI agent, who adapts the parameters of the virtual environment in real time based on the patient’s emotional and physiological state. This continuous flow of information enables adaptive, biofeedback-driven therapy and supports the ongoing refinement of BRAVE’s computational models, contributing to the long-term goal of creating intelligent systems capable of learning therapeutic strategies from clinical expertise. This thesis explores the development of the virtual environment and control infrastructure forming the experimental foundation of BRAVE’s therapeutic system. The environment focuses on aviophobia, one of the phobias targeted by the project, and has been designed to provide realistic, interactive and emotionally engaging exposure scenarios. Developed in Unity, the environment integrates custom 3D models created in Blender with a commercial aircraft template adapted for the project. Dynamic elements such as turbulence, lighting and sound are managed through C-Sharp scripting, ensuring immersive realism and flexible stimulus control. A web-based control panel, developed in JavaScript, allows therapists and researchers to manipulate exposure parameters in real time and monitor system states during sessions. This interactive dashboard communicates with the Unity environment through network protocols, forming a bidirectional connection that enables adaptive modulation of stimuli based on participants’ responses. The system is fully compatible with BRAVE’s Multimotion subsystem, a machine learning architecture that processes biosignals to predict patients’ emotional states in real time. This emotional monitoring assists therapists during sessions and will eventually train the AI-based Virtual Reality Exposure Therapy (AI-VRET) agent to autonomously adjust therapeutic strategies. The resulting system provides both a therapeutic tool and a research platform for data collection, model validation and AI training within BRAVE’s framework. In conclusion, this thesis contributes to the development of an AI-driven, adaptive Virtual Exposure Therapy ecosystem, establishing the technological foundation for future personalised and data-driven digital therapeutics applicable to the treatment of other phobias and affective disorders.