Embodied artificial intelligence as a paradigm shift for human–robot collaboration
0 
Abstract
Human–robot collaboration (HRC) has traditionally relied on instruction-driven paradigms in which humans specify goals and robots execute predefined tasks. Recent advances in embodied artificial intelligence (Embodied AI) challenge this model by grounding intelligence in physical embodiment and continuous interaction with the environment. This editorial positions Embodied AI as a paradigm shift in HRC, redefining collaboration as a physically interactive and mutually adaptive process. It examines the key challenges introduced by this shift and outlines emerging directions for future embodied HRC.
Keywords
1. INTRODUCTION
Human–robot collaboration (HRC) has long been envisioned as a future development of intelligent automation, with applications spanning manufacturing, healthcare, service robotics, and assistive technologies. During recent decades, remarkable progress has been achieved in motion planning, safety-aware control, intention recognition, and task allocation, enabling robots to collaborate with humans. However, most existing HRC systems remain fundamentally instruction-driven, where humans specify goals or constraints, and robots execute tasks within predefined operational envelopes.
Recently, embodied artificial intelligence (Embodied AI) has emerged as a growing research trend to emphasize the inseparable coupling between perception, action, and physical embodiment. Instead of treating artificial intelligence (AI) as a purely computational process separate from the physical world, Embodied AI posits that cognition arises through continuous interaction between the physical agent and the surrounding environment.
This editorial argues that Embodied AI is not merely an enabling technology for HRC, but a paradigm shift that redefines the nature of collaboration itself. Specifically, Embodied AI redefines HRC as a physically interactive process of mutual interaction in real-time, rather than a sequence of predefined task executions. This shift has profound implications for how robots are designed and how long-term human–robot partnerships are conceptualized.
2. CHALLENGES
As with any new technological paradigm, the application of Embodied AI to HRC entails both opportunities and challenges.
2.1. Modeling and learning mutual collaborative embodiments
A fundamental challenge lies in how to formally model and learn mutual embodiment between humans and robots. Most existing HRC systems remain robot-centric, with human behavior modeled as context rather than as a co-evolving component of the system. Mutual embodiment requires adaptiveness to human-to-robot and robot-to-human interactions across individual differences and contextual variability. From this perspective, challenges such as human intent inference, role negotiation, timing and turn-taking, and the development of team fluency are not separate issues but core requirements for realizing mutual collaborative embodiment. The difficulty of learning the co-adaptive representations is further exacerbated by the scarcity of high-quality human–robot interaction data and the persistent sim-to-real gap, where contact dynamics, human variability, and task diversity limit generalization across platforms and environments.
2.2. Integrating physical and cognitive human states
Although many studies have been conducted on the detection of human fatigue, workload, and affective states, these factors are still predominantly treated as external constraints or safety triggers. Embodied HRC demands a shift toward the deep integration of human physical and cognitive states into the control and learning loop, where such states actively shape robot behavior in real time. Beyond low-level physical states, effective embodied collaboration also requires continuous inference of higher-level cognitive states, including human intent, preferences, acceptable risk, and evolving task understanding. A major challenge is how to reconcile the high-level cognition with the stringent real-time control requirements of embodied interaction, where tight feedback loops involving force, compliance, and micro-adjustments are essential for fluent and safe collaboration.
2.3. Safety and responsibility in embodied collaboration
As AI becomes physically embodied, its decisions and actions increasingly have direct impacts on human well-being, and in some cases on safety and life itself, particularly in domains such as healthcare and autonomous driving. The physical embodiment of AI transforms decision-making from abstract computation into real-world action, raising fundamental questions about moral responsibility, accountability, and ethical governance. At the technical level, embodied HRC should ensure safety in physical interaction, where errors can lead to collisions, tool misuse, unsafe trajectories, or harmful forms of over-assistance in long-horizon tasks.
3. PERSPECTIVES
Embodied AI opens up new perspectives and possibilities for the evolution of HRC to encourage a shift toward more adaptive and human-centered forms of collaboration.
3.1. Humanoid embodied systems
Humanoid embodied systems represent a compelling pathway toward more transferable and scalable forms of HRC. By sharing similar body structures, sensorimotor capabilities, and action spaces with humans, humanoid robots offer a promising basis for learning from human demonstrations. This structural alignment may help narrow the gap between human intent and robotic execution, supporting collaborative behaviors that extend across tasks, environments, and application contexts. Future platforms should be well-suited to emerging architectures that combine embodied foundation models for semantic understanding and task decomposition with modular planning, verification, and low-level control stacks for contact-rich execution. In practice, embodied HRC could begin in structured and constrained environments, where the reliability and cost barriers of humanoid systems can be addressed before broader deployment in open-world settings.
3.2. Extension of human cognition and affect
Beyond task execution and adaptive control, Embodied AI invites a reconceptualization of artificial agents as extensions of human cognition and affect. By embedding intelligence within a physical body that continuously interacts with humans, Embodied AI systems can ground perception, decision-making, and interaction in meaningful social and emotional contexts. This enables robots to perceive and respond to affective signals such as engagement, stress, and trust. The future study could broaden the scope of HRC by positioning perception and sociality as embodied phenomena, while raising new questions regarding ethical responsibility and the societal role of emotionally responsive artificial agents. From this perspective, the shift from pre-programmed collaboration toward co-adaptive interaction reflects a move to shared autonomy, in which robots ask clarifying questions, negotiate roles, and learn user-specific conventions while keeping humans in meaningful control.
3.3. Trustworthy human–robot partnerships
Embodied AI might enable a transition from short-term task assistance to sustained human–robot partnerships. By grounding robot behavior in shared embodiment and continuous adaptation, collaboration can become more intuitive, resilient, and trust-aware. Human-robot partnerships will be particularly valuable in domains where human conditions, environments, and objectives evolve over time. Although the vision of trustworthy human–robot partnerships emphasizes long-term collaboration, real-world adoption is likely to progress unevenly, with earlier success in structured and constrained domains before broader deployment in open-world settings. This progression underscores the importance of reliability, transparency, and recoverability as foundational properties for trust-aware deployment rather than secondary design considerations.
4. CONCLUSION
Embodied AI marks a fundamental shift in how HRC is conceptualized, moving beyond instruction-driven task execution toward physically grounded and mutually adaptive interaction. This editorial has highlighted that while the embodied HRC introduces substantial challenges in modeling, learning, safety, and responsibility, it also opens promising perspectives for humanoid systems, shared cognition, and long-term human–robot partnerships. Importantly, the significance of embodied HRC is to enable collaboration that is adaptive, trustworthy, and centered on human needs and values. Realizing this vision will require rethinking system architectures, learning paradigms, and evaluation criteria across both research and deployment. Finally, Embodied AI invites the robotics community to reconsider not only how robots act, but how humans and robots coexist and collaborate in the physical world.
DECLARATIONS
Authors’ contributions
Worte the first draft: Li, J.
Made additions to the first draft: Yang, S. X.
Availability of data and materials
Not applicable.
AI and AI-assisted tools statement
Not applicable.
Financial support and sponsorship
None.
Conflicts of interest
Yang, S. X. is Editor-in-Chief of the journal Intelligence & Robotics, and Li, J. is a Junior Associate Chief Editor of the journal Intelligence & Robotics. Neither of them was involved in any steps of editorial processing, notably including reviewer selection, manuscript handling, and decision making, while the other authors have declared that they have no conflicts of interest.
Ethical approval and consent to participate
Not applicable.
Consent for publication
Not applicable.
Copyright
© The Author(s) 2026.
Cite This Article
Open Access
How to Cite
Download Citation
Export Citation File:
Type of Import
Tips on Downloading Citation
Citation Manager File Format
Type of Import
Direct Import: When the Direct Import option is selected (the default state), a dialogue box will give you the option to Save or Open the downloaded citation data. Choosing Open will either launch your citation manager or give you a choice of applications with which to use the metadata. The Save option saves the file locally for later use.
Indirect Import: When the Indirect Import option is selected, the metadata is displayed and may be copied and pasted as needed.
About This Article
Copyright
Data & Comments
Data
0
Comments
Comments must be written in English. Spam, offensive content, impersonation, and private information will not be permitted. If any comment is reported and identified as inappropriate content by OAE staff, the comment will be removed without notice. If you have any queries or need any help, please contact us at [email protected].