Affective Computing Meets Embodied Intelligence: Emotion Perception, Reasoning, and Interaction in Embodied Agents

Affective computing aims to endow machines with the ability to recognize, interpret, and respond to human emotions, while embodied intelligence emphasizes that cognition emerges through an agent's physical interaction with its environment. Historically, these two fields have advanced along largely separate trajectories: affective computing has focused on emotion recognition from facial, vocal, physiological, and textual signals, whereas embodied intelligence has concentrated on perception, control, and manipulation in the physical world. Recent breakthroughs in large multimodal models, foundation models for robotics, and human-robot interaction have created unprecedented opportunities to unify the two. Emotion is no longer treated as a passive signal to be classified, but as an active component that shapes how embodied agents perceive, reason, decide, and act within dynamic, socially rich environments.The goal of this research topic is to explore how affective computing can be deeply integrated into embodied intelligence to enable agents—such as social robots, service robots, virtual humans, and wearable systems—to understand and respond to emotions in a naturalistic and context-aware manner. Current embodied systems often perform well on physical tasks yet remain emotionally unaware, limiting their acceptance in real-world human-centered scenarios such as healthcare, education, elderly companionship, and collaborative work. We seek to bring together researchers from robotics, machine learning, affective science, and human-computer interaction to address key questions: How can embodied agents jointly perceive emotional and physical cues from multimodal streams? How should emotion inform action selection, motion planning, and social behavior? How can we build agents whose emotional expression is coherent with their embodiment? By consolidating recent advances and identifying open challenges, this collection aims to establish a shared research agenda for emotionally intelligent embodied systems and to promote trustworthy, empathetic, and effective human-agent collaboration.This research topic welcomes original research, reviews, methods, and perspective articles addressing the intersection of affective computing and embodied intelligence, including but not limited to: • multimodal emotion perception in embodied agents (vision, speech, physiological, and tactile signals); • emotion-aware decision-making, motion planning, and control; • affective human-robot and human-agent interaction; • emotional expression and generation in robots and virtual humans; • foundation and large multimodal models for embodied affect; • datasets, benchmarks, and evaluation protocols for embodied affective systems; • and applications in healthcare, education, elderly care, social companionship, and collaborative robotics. We also encourage contributions on ethical, trustworthy, and privacy-preserving design of emotionally intelligent embodied systems. Submissions may take the form of Original Research, Reviews, Mini-Reviews, Methods, Perspectives, or Data Reports. Topic editors: Fei Ma, Richard Yu, Weiwei Jiang Source: https://www.frontiersin.org/research-topics/84338/affective-computing-meets-embodied-intelligence-emotion-perception-reasoning-and-interaction-in-embodied-agents