An Optimal Control Framework for Adaptive Virtual Reality Interventions in Substance Use Rehabilitation
Year: 2026
Type: preprint
Abstract: Despite advances made over the years, substance use disorders are still a significant public health challenge, with relapse prevention continuing to be a major problem in rehabilitation. Traditional rehabilitation strategies often face barriers such as limited accessibility and difficulty recreating real-world environments that are associated with cravings. As a result, researchers have begun to study virtual reality (VR) as a tool for drug rehabilitation. I developed an Adaptive-Perturbed-Corrective-Relapse (APCR) compartmental model governed by differential equations that represent the transitions between stable recovery, cue-induced perturbation, corrective intervention, and relapse. The model incorporates a craving trigger parameterized using alcohol-related gaze data collected in a VR environment and a baseline recovery cue derived from physiological biomarkers. Transition rates were assigned based on literature-informed assumptions rather than clinical outcome data. To determine when and how strongly therapeutic interventions should be delivered, I formulated an optimal control problem in which intervention intensity is represented by a time-dependent control variable. Using Pontryagin’s Minimum Principle, I derived an intervention strategy that balances reducing the perturbed state with the control effort over 60 minutes. Using the forward–backward sweep method, I obtained an optimal control strategy that reduced the simulated peak proportion of individuals in the Perturbed state by 33.26% compared with the passive baseline, highlighting the potential of mathematical optimization for adaptive VR rehabilitation. This work supports the United Nations Sustainable Development Goal 3 (Good Health and Wellbeing).
Source: Research Square
Authors: Charlotte Cheng, Padmanabhan Seshaiyer
Institution: George Mason University
Language: English
Subfield: Epidemiology
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7226020443
