Human-Centered Vision-Language-Action Models for Embodied Interaction in Augmented, Virtual, Mixed, and Smart Environments

Human-media interaction research is undergoing a transformative shift, driven by the convergence of multimodal perception, natural language processing, and embodied intelligence. Unlike models that focus primarily on perception, reasoning, and response generation, Vision-Language-Action (VLA) models are artificial intelligence systems that integrate visual perception, language understanding, and action generation. Their ability to interpret users, understand the surrounding environment, and generate purposeful actions has the potential to transform human–media interaction from passive information exchange into active, context-aware collaboration. This opens new opportunities in augmented, virtual, and mixed reality, embodied agents, assistive robotics, and smart environments.However, despite rapid advances in VLA models and embodied AI systems, significant challenges remain in grounding these systems in human-centered principles that ensure trust, safety, transparency, usability, and inclusivity. Current approaches often emphasize model capability, task completion, or benchmark performance, while less attention has been given to how users interact with, supervise, correct, and rely on these systems in real-world contexts. Without addressing these challenges, improvements in algorithmic performance may not translate into systems that users can effectively adopt. Bridging the gap between model capability and meaningful human experience remains an open research challenge requiring interdisciplinary investigation across artificial intelligence, human-computer interaction, cognitive science, design, and extended reality.This Research Topic aims to explore the design, development, and evaluation of human-centered VLA models for embodied interaction in augmented, virtual, mixed, and smart environments. The primary goal is to advance understanding of how VLA-based agents can perceive, communicate, reason, and act collaboratively with humans to support real-time, adaptive, and context-aware interaction. By foregrounding human experience, this Research Topic encourages contributions that address usability, trust, ethics, accessibility, and social acceptance in VLA-enabled embodied AI systems. Authors are invited to investigate how users can instruct, monitor, correct, or share control with action-capable systems; how these systems can explain uncertainty, intent, and planned actions; and how they can be effectively adapted for diverse domains, from assistive and educational technologies to industrial and healthcare applications.To gather further insights into the design, implementation, and assessment of human-centered VLA systems for embodied interaction, we welcome articles addressing, but not limited to, the following themes:- Human-centered design and evaluation methods for VLA-based systems- VLA models in augmented, virtual, and mixed-reality environments- Interactive, embodied, and socially aware agents with perception, dialogue, and action capabilities- Human-in-the-loop supervision, shared autonomy, and adaptive assistance mechanisms- Ethical, transparent, and explainable action-generation strategies for embodied agents- Assistive, educational, and accessible applications of embodied VLA models- Smart environments, wearable and tangible interfaces supporting collaborative action- Evaluation frameworks integrating task success with trust, comfort, and safety metrics- Datasets, simulation tools, and reproducible protocols for studying embodied VLA systems- Governance, privacy, and long-term implications of action-capable AI in human contexts Topic editors: Akshith Ullal, Zhao Han, Xiangyi Cheng, Chen Chen Source: https://www.frontiersin.org/research-topics/84456/human-centered-vision-language-action-models-for-embodied-interaction-in-augmented-virtual-mixed-and-smart-environments