Physiological sensing for situational awareness: a theory-driven integrative review of multimodal and unsupervised...

Situational awareness (SA) is fundamental to performance in visually demanding, safety-critical tasks and in human-autonomy teaming (HAT), yet existing measurement approaches remain limited. Direct probes such as SAGAT or SART disrupt task flow, while performance-based metrics capture outcomes rather than the underlying cognitive processes of perception, comprehension, and projection. This article presents a theory-driven integrative review of physiological sensing approaches for estimating SA. Physiological sensing provides a complementary, continuous, and objective alternative, with evidence that neural (EEG), ocular (eye tracking and pupillometry), and autonomic (ECG and EDA) signals each index distinct components of SA. The review synthesizes findings across these modalities with emphasis on two key domains: visual search tasks, where operators must detect, classify, and assess potential threats, and human-autonomy teaming, where effective coordination depends on shared SA, trust, and transparency. Beyond traditional feature-based pipelines, we examine recent advances in multimodal fusion and deep learning, highlighting the increasing role of unsupervised and self-supervised representation learning. These approaches exploit large volumes of unlabeled physiological data to reveal latent cross-modal structure, reduce reliance on sparse ground-truth labels, and enhance scalability and ecological validity. By integrating evidence across sensing modalities and computational frameworks, this review outlines the current state of physiological SA estimation and identifies key research directions for continuous, real-time monitoring in visual search and human-autonomy teaming.