Effective firefighter training demands practice under dynamic, high‑stress emergencies, yet live‑fire drills are expensive, risky, and hard to scale. Virtual Reality (VR) offers a safe, controllable alternative, but most existing VR firefighter systems rely on fixed scenarios or provide only limited difficulty adaptation, making it difficult to keep the challenge level aligned with trainees' evolving performance. To overcome this limitation, we introduce ProACT, a performance‑driven adaptive training framework that closes the loop between trainee performance and simulated fire dynamics. ProACT consists of two modules: a global module that continuously adjusts scene‑level fire spread speed in real time based on performance metrics, and a local module that periodically updates object‑level flammability across the environment. Evaluation shows that ProACT outperforms both adaptive and non‑adaptive baselines in objective training outcomes and subjective user experience. We also discuss how adaptive immersive environments can be extended beyond firefighting to other training domains.
Review