Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition characterized by challenges in communication, emotional regulation, and sensory processing, making individuals with ASD prone to elevated, often undetected stress. Timely stress detection is critical for early intervention and improved behavioral outcomes. This study reviews recent advancements in wearable sensors and machine learning (ML) techniques for real-time stress monitoring in ASD. Key approaches include physiological signal acquisition heart rate (HR), electrodermal activity (EDA), skin temperature (ST), and motion analyzed using models such as SVM, Random Forest, CNN, and fuzzy logic. The integration of wearable platforms, embedded systems, and mobile health applications enables haptic feedback, real-time alerts, and remote caregiver monitoring. Despite limited datasets, findings highlight the potential of combining intelligent systems with physiological monitoring to support personalized ASD interventions. Challenges related to data diversity, algorithm generalization, hardware usability, and clinical deployment are identified, and future directions for robust, context-aware stress detection are discussed.