Gait disorders in Parkinson's disease (PD), particularly in complex environments or under stress, present challenges for accurate evaluation and classification, such as in cases of freezing of gait. Traditional clinical and laboratory settings often fail to replicate the complexity needed for precise classification, making effective rehabilitation difficult. This study aims to address these challenges by developing an augmented reality (AR)-based environment that mimics real-world stressors and dynamically adapts to the patient's condition. The AR system is designed to facilitate individualized gait training and rehabilitation by modifying environmental difficulty based on real-time feedback from gait performance and stress levels. Building on Gentile's taxonomy of tasks, the investigators have incorporated PD-specific factors, such as cognitive dual tasks, into our environment classification system. Preliminary results suggest that this system effectively elicits varying gait and heart rate variability (HRV) responses, indicating different stress levels. This trial will further test the AR environment's ability to classify patients based on their responses to complex, interactive environments, while also investigating the effects of adaptive AR-based gait training on both gait and stress management in individuals with PD.