Embodied Intelligence Networks (EIN)
A half-day workshop on networking for embodied and agentic intelligence, bringing together researchers working on perception, reasoning, and action across virtual and physical environments.
Overview
AI, as Andrej Karpathy referred to in his tweet last January as “IA (intelligence amplification)”, is more than just independent long-running agents. It increasingly has the vibe of tools for thought, needing human interaction. Karpathy’s vision has quickly become a reality of Embodied Intelligence, where autonomous agents are increasingly moving from virtual environments into the physical world, interacting with humans, encompassing physical entities such as collaborative robots, smart glasses, and industrial IoT sensors. These autonomous agents are capable of real-time perception of their surroundings, and can take actions—but both their perception and their decisions critically depend on an efficient, low-latency, highly dependable, and adaptive network infrastructure.
In both virtual and physical environments, with the ubiquity of tool use (e.g., via the Model Context Protocol), the ways how autonomous agents interact and network with one another will inevitably evolve beyond traditional data transport and connectivity, and become an embodied intelligence network with agentic workflows. Such a network resembles an intelligent mesh consisting of actors that communicate with each other by message passing, running agents who are capable of perceiving and adapting to the environment, understanding human intent, coordinating multi-agent activities, and providing semantic communication with real-time control loops.
This new workshop aims to focus squarely on exciting ideas in the emerging topic of agentic computing, as well as how networking can adapt to the needs of agentic computing, in both virtual and phyiscal environments. We encourage submissions of forward-looking research, novel system designs, and practical deployment experiences that deeply integrate networking with embodied perception, reasoning, and action. Papers with open-source artifacts — including source code and datasets — are especially welcome and strongly encouraged. The freshness of visionary and forward-looking ideas is more important than the completeness of theoretical proofs, and simplicity may not be inferior to complexity.
Important Dates
- Submission deadline
- December 29, 2025
- Notification of acceptance
- February 2, 2026
- Camera-ready manuscript due
- February 16, 2026
- Workshop Date
- May 18, 2026
- Workshop format
- Half-day event held in conjunction with IEEE INFOCOM 2026
Dates may be adjusted to follow the final INFOCOM 2026 workshop schedule; please check back for updates.
Scope
EIN sits at the intersection of agentic computing and next-generation networking. It asks how networks should evolve as embodied intelligence becomes mainstream and agents become ubiquitous — across edge, cloud, and physical environments.
The workshop builds on prior work in distributed and federated learning, but focuses specifically on the networking challenges of real-time control loops, ultra-low-latency coordination, semantic communication, and multi-agent orchestration for embodied systems.