Workgroup:Network Working Group                                     X.Zhang
Internet-Draft:                                                    Hicagent
draft-zhang-rvp-problem-statement-00                                Y.Zhang
Published:20 October 2025                                           C.Zhang
Intended status: Informational                                      M.Zhang
Expires:20 April 2026                                         China Telecom
                                                                       W.Lu
                                                                     SUPCON




    Problem Statements and Requirements of Real-Virtual Agent Protocol
   (RVP): Communication Protocol for Embodied Intelligence in Physical-
                            Digital Continuum





    Status of this Memo

  This Internet-Draft is submitted in full conformance with the
  provisions of BCP 78 and BCP 79. This document may not be modified,
  and derivative works of it may not be created, except to publish it
  as an RFC and to translate it into languages other than English.

  This document may contain material from IETF Documents or IETF
  Contributions published or made publicly available before November
  10, 2008. The person(s) controlling the copyright in some of this
  material may not have granted the IETF Trust the right to allow
  modifications of such material outside the IETF Standards Process.
  Without obtaining an adequate license from the person(s) controlling
  the copyright in such materials, this document may not be modified
  outside the IETF Standards Process, and derivative works of it may
  not be created outside the IETF Standards Process, except to format
  it for publication as an RFC or to translate it into languages other
  than English.

  Internet-Drafts are working documents of the Internet Engineering
  Task Force (IETF), its areas, and its working groups.  Note that
  other groups may also distribute working documents as Internet-
  Drafts.

  Internet-Drafts are draft documents valid for a maximum of six
  months and may be updated, replaced, or obsoleted by other documents
  at any time.  It is inappropriate to use Internet-Drafts as reference
  material or to cite them other than as "work in progress."



    Zhang                  Expires 20,April,2026                 [Page 1]

    Internet-Draft          RVP Problem Statement             October 2025


  The list of current Internet-Drafts can be accessed at
  http://www.ietf.org/ietf/1id-abstracts.txt

  The list of Internet-Draft Shadow Directories can be accessed at
  http://www.ietf.org/shadow.html

  This Internet-Draft will expire on April 20, 2026.

    Copyright Notice

  Copyright (c) 2025 IETF Trust and the persons identified as the
  document authors. All rights reserved.

  This document is subject to BCP 78 and the IETF Trust's Legal
  Provisions Relating to IETF Documents
  (http://trustee.ietf.org/license-info) in effect on the date of
  publication of this document. Please review these documents
  carefully, as they describe your rights and restrictions with
  respect to this document. Code Components extracted from this
  document must include Simplified BSD License text as described in
  Section 4.e of the Trust Legal Provisions and are provided without
  warranty as described in the Simplified BSD License.

Abstract

  The Real-Virtual Agent Protocol (RVP) enables seamless coordination
  between physical entities (robots, IoT devices, manufacturing
  systems and agents) and digital agents (AI systems, software agents,
  virtual twins) through unified composite identity management,
  physical/social/production relations graph-based coordination, and
  physical constraint integration. Unlike existing protocols that
  assume peer-to-peer digital relationships (A2A for agents, MCP for
  AI tools, ANP for agent networks), RVP unifies physical and digital
  agents communication and achieves physical data loop for online
  learning for embodied agents considering both hierarchical
  physical/social/production relations and physical world constraints.
  RVP is designed for immediate deployment in modern manufacturing,
  smart cities, autonomous mobility systems, and human-AI
  collaborative environments where non-peer, partially centralized
  relations and coordination is essential for real-world embodied
  intelligence networks.








    zhang                  Expires 20,April,2026                 [Page 2]

    Internet-Draft          RVP Problem Statement             October 2025


    Table of Contents


1. Introduction...........  .....................................  ...3
    1.1. Background and Current Status of Existing Protocols....   ...3
    1.2. Why RVP is Necessary....................................   ..6
2. Terminology...  ...............................................  ..7
3. Problem Statement...  ..........................................  .8
     3.1. Identity Fragmentation Across Domains....................  .8
     3.2. Lack of Physical/Social Relations Mapping among agents...  .8
     3.3. Lack of Physical Constraint Reflection..................  .10
     3.4. Lack of Efficient Registration and Effective Discovery       
     Mechanisms for Agents........................................ . 10
     3.5. Lack of Real-time Data Input to the Embodied Large Model for 
     Training.....................................................  .10
4. Protocol Requirements Discussion and Design Principles...  ....  .11
     4.1. Design Principles..................................... . ..11
     4.2. Requirements Discussion................................  ..12
     4.3. Protocol Integration Strategy..........................  ..15
5. RVP Use Cases......  ..........................................  .17
     5.1. Flexible Production Line Coordination in Manufacturing..  .17
     5.2. Disaster Rescue Human-Robot Collaborative Environments..  .18
     5.3. Autonomous Driving Systems in Smart City................  .18
        5.3.1. Multi-Vehicles(Agents) Collaborative Delivery..... .. 18
        5.3.2. Vehicle-to-Everything (V2X) Intersection Management.  19
6. Security Considerations........................................  .19
     6.1. Identity Verification and Authorization....................19
     6.2. Communication Security.....................................19
     6.3. Safety System Protection...................................20
     6.4. Audit and Monitoring.......................................20
     6.5. Isolation Capabilities.....................................21
7. IANA Considerations............................................  .21
8. References.....................................................  .21
     8.1. Normative References.......................................21
     8.2. Informative References.....................................21
9. Acknowledgments................................................  .22


1. Introduction
1.1. Background and Current Status of Existing Protocols

  The convergence of artificial intelligence, robotics, and ubiquitous
  computing is creating unprecedented demand for connectivity and
  coordination between physical systems and digital intelligence.
  However, existing protocols are insufficient to effectively meet
  these demands, facing severe challenges in cross-entity
  collaboration, cross-level management, and cross-architecture


    zhang                  Expires 20,April,2026                 [Page 3]

    Internet-Draft          RVP Problem Statement             October 2025


  communication, primarily manifested in problems such as identity
  fragmentation, protocol incompatibility, and missing physical
  constraint reflection mechanisms.

  With the rapid development of Large Language Models (LLMs), LLM-
  based agents have become very popular in many industries and
  scenarios. Agent and model protocols are gaining widespread
  attention from both industry and academia. Meanwhile physical
  agents(e.g., emobodied robots, autonomous driving cars, UAVs)
  embedded in physical entities are playing a crucial role in modern
  manufacturing[SMART-FACTORY], smart cities[AI-CITY], autonomous
  mobility systems[APPOLO], and human-AI collaborative
  environments[Human-Robot]. The interaction between hetergenous
  physcial agents and digital agents as well as models bridges the
  physical world and digital world into a seamless continuum. Real-
  Virtual Agent Protocol (RVP) is a communication protocol designed
  for reality-virtuality symbiotic, supporting non-peer-to-peer, and
  partially centralized embodied intelligence agent networks. It
  focuses on mapping on physical/social/production relations graphs
  and physical constraint reflection between agents.

  Today's agent communication protocols are designed for homogeneous
  environments. Mainstream agent communication protocols include
  A2A[A2A-SPEC], MCP[MCP-SPEC], ANP[ANP-SPEC], etc.:

  Agent-to-Agent (A2A) Communication Protocol: Supports digital agent
  communication but assumes all participants are software entities
  operating in digital environments in peer-to-peer communication,
  hardly to handle non-peer relationships in physical world, and lacks
  of physical/social/production relations constraints. In practical
  applications, it has fundamental limitations for physical-digital
  coordination, including:

  1) Communication Model Mismatch: A2A protocol uses asynchronous
  message transmission optimized for digital agents that can pause,
  queue, and process messages at variable speeds. Physical systems
  require real-time coordination with hard deadlines-a robot cannot
  "pause" mid-motion to wait for queued messages.

  2) No Physical Constraint Modeling: A2A protocol has no standard way
  to represent physical constraints (workspace boundaries, inertia,
  safety requirements). When a digital agent requests a robot to "move
  to position X immediately," the protocol cannot express that the
  robot needs 3.2 seconds due to physical acceleration limits.





    zhang                  Expires 20,April,2026                 [Page 4]

    Internet-Draft          RVP Problem Statement             October 2025


  3) Safety Integration Gap: A2A protocol has no built-in safety
  mechanisms. There is no standard way to represent that certain agent
  communications could cause physical harm if mishandled.

  MCP: Provides AI systems with tool access but treats physical
  systems as external, and sometimes stateless tools rather than
  intelligent twin participants. It focuses on model context
  management, concentrating on context interaction between agents and
  tools (such as user preferences, environmental states, session and
  task information, etc.). The problems raised by physcial systems
  include:

  1)Stateless Tool Model: MCP treats all tools as stateless functions.
     Physical systems are inherently stateful, i.e., a robot's current
     position affects what next operations are possible. MCP is hard to
     represent that the same tool call may succeed or fail based on
     physical context.

  2)Bidirectional Coordination: MCP is bascially request-response model.
     Physical systems need to provide continuous feedback (sensor data,
     status updates, emergency conditions) that should influence AI
     decision-making in real-time.

  3) MCP basically assumes one single AI system/agent accessing tools,
     which does not involve complex collaboration relationships between
     agents. Physical agents often need to coordinate between multiple
     AI systems (planning AI, safety monitoring AI, quality control AI)
     simultaneously. Moreover, real-time interaction and
     synchronization (e.g., ms magnitude) between physical and digital
     agents is hard to achieve.


  Agent Network Protocol (ANP): Organizes agent networks in multi-
  agent scenarios with peer-to-peer entities, building the underlying
  architecture of the agent internet, known as the "HTTP of P2P
  networks" and supporting decentralized decision-making. It supports
  decentralized discovery but identity management is based on DID,
  unable to handle 1-to-N topology relations from the physical entity
  to virtual entities (for examaple, an automonous driving car
  involves a physcial entitiy and several virtual assistant entities
  to do route optimization and simulation, operate the on-car
  electronic systems and search for the weather forecast,etc.)and
  virtual entity cluster discovery requires multiple queries with low
  efficiency. Its constraints are similar to the A2A communication
  protocol and will not be elaborated further.


    zhang                  Expires 20,April,2026                 [Page 5]

    Internet-Draft          RVP Problem Statement             October 2025


  Additionally, regarding connecting physical and digital entities,
  existing underlying transport protocols have the following issues:

  IoT Protocols (MQTT, CoAP[RFC7252], HTTP) handle device connectivity
  and data exchange but lack semantic understanding of physical-
  digital integration requirements.

  IoT protocols like MQTT and CoAP handle device connectivity but
  cannot support intelligent coordination:

  1) No Coordination Semantics: IoT protocols transport data but have
  no understanding of coordination requirements. Publishing sensor
  data to a topic provides no information about what coordination
  actions are appropriate.

  2) No Identity Coherence: IoT protocols identify devices by topics
  or endpoints but cannot represent that multiple endpoints refer to
  the same logical entity. A robot's position sensor, control
  interface, and status feed are separate MQTT topics with no
  protocol-level connection.

  3) No Intelligence Integration: IoT protocols have no standard way
  to integrate with AI systems beyond basic pub/sub. There is no
  protocol support for AI systems to understand device capabilities or
  coordinate intelligent actions.

  RVP is implemented similarly to existing protocols (such as MCP, A2A,
  ANP) through further encapsulation or modification of underlying
  transport protocols (such as SSE, WebSocket, MQTT) or even direct
  modification of A2A, MCP and ANP.

1.2. Why RVP is Necessary

  Existing protocols (A2A, MCP, IoT protocols) have fundamental
  limitations in handling unified coordination of physical entities
  and digital entities. Although A2A protocols could theoretically be
  extended to handle physical entities, this would require fundamental
  changes to agent communication semantics to handle physical
  constraints [A2A-FIPA][CPS-DESIGN]and complete redesign of unified
  identity management across domains. Enhancing MCP to effectively
  handle physical and digital agent coordination would require new
  design principles to design stateful coordination and real-time
  communication among different kinds of agents, which is common in
  real world. It's hard to imagine to convert all physical systems
  into stateless tools, thereby losing their needs of intelligent
  coordination. Adding intelligence to IoT protocols would require
  fundamental semantic changes incompatible with existing IoT


    zhang                  Expires 20,April,2026                 [Page 6]

    Internet-Draft          RVP Problem Statement             October 2025


  deployments, still lacking unified identity management and limited
  to device-level rather than system-level coordination.

  Future multi-agent architectures will evolve into hierarchical
  (MCP's C/S architecture can satisfy), distributed (ANP's P2P
  architecture can satisfy), and hybrid frameworks, enabling flexible
  and diverse ways to coordinate large numbers of agents to complete
  various complex tasks under different business scenarios. However,
  the existing architectures cannot fully meet the communication needs
  of reality-virtuality integrated embodied intelligence agent
  networks. For example, one-to-many real-virtual mapping requires
  simultaneous identity authentication of multiple virtual entities
  and simultaneous registration and discovery of these agents.
  Communication between real-virtual agents also requires following
  constraints and rules of production relations graphs in the physical
  world. What's more, the data starvation for model training requires
  real-time physical data input to achieve data-loop for online
  training in embodied intelligence.

  Therefore, RVP is needed that can treat physical and digital as
  expressions of unified entities and create efficient communication
  based on these expressions. Instead of replacing existing protocols,
  the new protocol is designed to integrate with them:

  1) Integration with Existing Protocols: Through adapters or
  middleware layers, enabling the new protocol to work collaboratively
  with existing protocols such as A2A, MCP, IoT, etc.

  2) Introducing New Mechanisms: Particularly physical constraint
  reflection, unified identity management, physical/social/production
  relations graphs, etc. Based on this, a heterogeneous communication
  protocol for reality-virtuality integrated embodied intelligence
  agents is proposed.

2. Terminology

  The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
  "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
  document are to be interpreted as described in RFC 2119 [RFC2119].

  Composite Agent: A Composite Agent is a super-agent to interact with
  the external physical world on behalf of its associated Real-entity
  Agent, and Virtual-entity Agent. While adhering to the principle of
  permission isolation, it realizes effective utilization and
  comprehensive fusion of information related to both worlds. By
  abstractly layering the capabilities of individual entities in the
  physical world, it achieves efficient coordination from value


    zhang                  Expires 20,April,2026                 [Page 7]

    Internet-Draft          RVP Problem Statement             October 2025


  decision-making to task execution. Composite Agent aims to
  characterize human/unmaned behavior patterns in complex environments,
  ensuring efficiency, flexibility, and stability of task processing.

  Physical Expression: The concrete instance of an object in the
  physical domain, carrying the existence form and interaction
  capabilities of the object in the physical world.

  Digital Expression: The virtualized instance of an object in the
  digital domain, carrying the logical representation and
  computational capabilities of the object in digital space.

  Real-Virtual Agent Protocol (RVP): A standardized communication
  protocol specifically designed for composite agent architecture,
  regulating interactions and coordination among physical and digital
  agents.

  Physical/Social/Production Relations Graph: A semantic network-based
  directed graph describing organizational relationships, business
  processes, and permission constraints among agents.

  Physical Constraint: Standardized description of timing, spatial,
  dynamic, and safety constraints in the physical world, supporting
  automatic verification and real-time reflection.

3. Problem Statement

3.1. Identity Fragmentation Across Domains

  Currently, in above mainstream communication protocols, no protocol
  provides unified identity management across physical and digital
  domains. Each protocol treats the same entity as a separate,
  unrelated object. The same logical entity exists as disconnected
  representations in different systems. These representations are
  manually coordinated through custom integration code, leading to
  consistency problems and coordination failures.

  Without unified identity management, the same entity has multiple
  un-associated identity representations in different systems, leading
  to low registration efficiency, data inconsistency, and coordination
  failures. Identity information needs to be manually synchronized,
  increasing system complexity and maintenance costs.

3.2. Lack of Physical/Social Relations Mapping among agents

  Current mainstream agent protocols (MCP/A2A/ANP) are built on a
  digital-native worldview, ignoring the inherent non-peer nature,


    zhang                  Expires 20,April,2026                 [Page 8]

    Internet-Draft          RVP Problem Statement             October 2025


  hierarchical structures, and business constraints of agents in the
  physical world. In existing agent systems, interactions between
  agents are often based on simple peer-to-peer or client-server
  models, lacking mapping of complex production relationships in the
  real world.

  In real world, an agent may represent the individual person's social
  roles in different scenarios, such as father, employee, friend, etc.;
  it can also represent actual existence in the physical world, such
  as robots, sensors, actuators, and other hardware devices.
  Collaboration in the real world is often non-peer, with hierarchical
  structures, division of labor, and permission constraints. For
  example, in intelligent manufacturing, there are levels such as
  scheduling centers, production line control, and robots; in smart
  cities, there are command centers, departments, and field devices.

  Therefore unlike in digital world, the hierarchical and network-
  based physical/social/production relations graph exists, which
  stores physical world data and capabilities related to specific
  roles.

  Due to the lack of explicit modeling and mapping of these
  relationships, it leads to:

    Difficult Permission Control: Access and operation permissions
  between agents are not constrained based on
  physical/social/production relations, potentially causing
  unauthorized operations.

    Low Collaboration Efficiency: Agents cannot quickly identify
  collaboration partners, requiring complex negotiation and discovery
  mechanisms.

    Unreasonable Resource Allocation: Due to the lack of relations-
  based scheduling, resource allocation may not conform to actual
  business logic, leading to resource waste or bottlenecks.

    Poor System Maintainability: When business relations change,
  manual adjustments to connections and permissions between agents are
  needed, which is error-prone.

  Therefore, we need to introduce a "Phycial/Social/Production
  Relations Graph" in the protocol to explicitly describe these
  relations among agents and implement permission control,
  collaboration scheduling, and resource allocation based on this
  graph.



    zhang                  Expires 20,April,2026                 [Page 9]

    Internet-Draft          RVP Problem Statement             October 2025


3.3. Lack of Physical Constraint Reflection

  The current agent protocol ecosystem has a fundamental cognitive
  disconnect: intelligent protocols like A2A and MCP can understand
  intelligent coordination but lack physical perception. In a research
  [STI-BENCH] it pointed out While LLMs have been extensively studied
  for visual semantic understanding tasks, their ability to perform
  precise and quantitative spatial-temporal understanding in real-
  world applications is rather week, with top-performing models like
  Gemini-2.5-Pro achieving around 41.4% average accuracy.

  IoT protocols can understand physical devices but lack intelligent
  cognition. This causes agent decisions in the digital world to be
  seriously disconnected from physical world reality rules.

  Physical and digital spaces need to establish a standardized
  description framework for physical constraints to implement real-
  time reflection mechanisms for constraint states, ensuring
  consistency between digital decisions and physical reality.

3.4. Lack of Efficient Registration and Effective Discovery Mechanisms
   for Agents

  Existing protocols (such as MCP, A2A, ANP) have deficiencies in
  registration and discovery, resulting in agents requiring manual
  configuration and inability to dynamically adapt to network changes.
  Due to the lack of unified, adaptive, semantic registration and
  discovery mechanisms, agent networks become information silos,
  unable to achieve true "plug-and-play" and "dynamic collaboration."
  A unified registration and discovery mechanism needs to be designed
  to support dynamic joining and exiting of agents and automatic
  updating of network topology.

3.5. Lack of Real-time Data Input to the Embodied Large Model for
   Training

  Back to 2023, just after one year of ChatGPT's release, many studies
  warned that the world could run out of high-quality data to train AI
  soon. The research[DATA-RUNOUT] predicted that the training dataset
  will run out between 2026 and 2032, or even earlier.

  This is even worse for embodied intelligence systems, which has less
  recored knowledge to learn. Online learning from real-time practice
  data attracts more and more attentions to solve this problem.
  Particularly, data automatically generated during agent task
  execution that is not directly explicitly recorded contains deep
  information about environmental dynamics, physical laws, task


    zhang                  Expires 20,April,2026                [Page 10]

    Internet-Draft          RVP Problem Statement             October 2025


  execution context, decision-making processes, and multi-modal
  interactions. For example, in sanitation scenarios, implicit data
  includes interactions between sanitation vehicles and the
  environment (such as the relationship between cleaning force and
  ground material), micro-adjustments during task execution (such as
  trajectory corrections when avoiding pedestrians), and correlations
  between equipment operating status and work effectiveness. These
  real-time data supplement is vital for training the embodied agents.
  Moreover it will help to solve the data-training-decision
  fragmentation at the same time, i.e., when training data is
  historical, static, and disconnected, while agents need to make
  immediate responses in dynamic, real-time, continuous environments.
  This fragmentation prevents models from continuously evolving in the
  real world.

  To the best of our knowledge, current protocols are not available to
  input real-time data for continuous training.

  To summary, there is an urgent need for unified communication
  protocol standards to effectively connect physical entities and
  digital agents.

4. Protocol Requirements Discussion and Design Principles

4.1. Design Principles

  Constraint Priority Principle: Physical constraints are not
  afterthought limitations but prerequisites for system design and
  operation.

    Perform physical feasibility verification before all decisions
  and executions

    Establish automatic propagation and consistency maintenance of
  constraints in task chains

    Deeply integrate safety constraints into coordination semantics

  Unified Semantics Principle: Establish a unified semantic framework
  across all protocols and domains, eliminating ambiguity in concept
  mapping.

    Ensure consistent interpretation of the same concepts across
  different protocols and domains

    Semantic interpretation considering current physical environment
  and business context


    zhang                  Expires 20,April,2026                [Page 11]

    Internet-Draft          RVP Problem Statement             October 2025


    Support runtime semantic adjustment and optimization

4.2. Requirements Discussion

  This section is intended to stimulate the open discussion for the
  RVP requirements, so that the design team can consolidate the
  requirements of the RVP protocol.

  RVP.REQ-1: Unified Entity Identity Management

  Addressing the identity fragmentation problem in existing protocol
  ecosystems, where the same entity has different identity marking
  systems in physical, digital, and hybrid domains, leading to
  semantic inconsistency, life cycle disconnection, and accumulated
  security risks. The protocol MUST support unified entity
  registration, containing physical capabilities, digital intelligence,
  and hybrid manifestations within a single identity model.

  The requirements include:

    Support automatic association and verification of physical and
  digital identities

    Establish a globally unique entity identifier system

    Identity registration MUST contain complete composite identity
  description.

    Identity verification SHOULD support cross-domain identity
  consistency checks.

    Identity updates MUST ensure synchronization of identity
  information in all domains

  RVP.REQ-2: Dynamic Registration and Discovery Mechanisms

  Existing systems rely on manual configuration and hard-coded
  connections, unable to adapt to dynamic network topology changes,
  resulting in poor system scalability. The protocol MUST provide
  efficient agent registration and service discovery mechanisms.

  The requirements include:

    New nodes MUST be able to automatically discover the network and
  complete registration




    zhang                  Expires 20,April,2026                [Page 12]

    Internet-Draft          RVP Problem Statement             October 2025


    Node capability changes MUST be updated to the registry in real-
  time

    Provide real-time node health status monitoring

    Support external interface for natural language semantic service
  discovery

    Service discovery MUST consider current context and environmental
  state

    Provide intelligent service selection based on multi-dimensional
  matching

  RVP.REQ-3: Physical/Social/Production Relations Graph Management

  Existing protocols lack of expressive capability for real-world
  organizational structures, unable to express business relations such
  as command, collaboration, and affiliation, resulting in low
  coordination efficiency and coarse-grained permission control. The
  protocol MUST provide standard mechanisms to define, maintain, and
  query organizational structures among entities based on production
  relationships.

  The Graph Model Requirements include:

    Define standardized physcial/social/production relations
  classifications

    Support runtime creation, update, and deletion of relations

    Implement automatic derivation of permissions based on relations
  graphs

  The Operational Requirements include:

    Node registration MUST automatically derive initial
  physical/social/production relations

    Relations changes SHOULD trigger impact analysis and system
  adjustments

    Routing decisions MUST consider physical/social/production
  relations constraints

  RVP.REQ-4: Physical Constraint Integration



    zhang                  Expires 20,April,2026                [Page 13]

    Internet-Draft          RVP Problem Statement             October 2025


  Existing protocols lack systematic description and verification
  mechanisms for timing, spatial, dynamic, and other constraints,
  causing digital decisions to be disconnected from physical reality.
  The protocol MUST provide standard mechanisms to represent and
  enforce physical constraints in coordination decisions.

  The requirements include:

    Define machine-readable physical constraint description format,
  forming a standardized constraint language

    Establish complete constraint classification

    Constraint description MUST support automatic verification

    Decision verification MUST perform physical constraint checks
  before execution

    Constraint violations SHOULD trigger standardized emergency
  responses

    Constraint propagation MUST maintain consistency during task
  decomposition

  RVP.REQ-5: Safety-Aware Coordination Semantics

  Safety mechanisms considerations are required to be added as built-
  in features to meet real-time requirements. The protocol MUST
  integrate physical safety constraints, timing requirements, and
  emergency response mechanisms into coordination communications.

  The requirements include:

    Built-in safety semantics, with safety constraints as core
  components of coordination semantics

    Safety decisions consider physical state and business context

    Operation execution SHOULD perform real-time safety boundary
  checks

    Emergency response MUST be automatically triggered when safety
  threats are detected

  RVP.REQ-6: Real-time Data Sharing




    zhang                  Expires 20,April,2026                [Page 14]

    Internet-Draft          RVP Problem Statement             October 2025


  Real-time data sharing involve injecting real-time data of the
  physical agents into models and sharing direct and local sense,
  observation or other kind of data between agents.

  The requirements include:

       Extremely Low latency reliable data transmission

       Streamed training data provision to training procedure

       Training and reasoning mode switch dynamically

       Model update and versioned distribution

       Standard gradient/parameter aggregation interface

       Direct sense and information sharing between agents

4.3. Protocol Integration Strategy

  RVP is designed as a meta-protocol that does not replace existing
  protocols (such as A2A, MCP, IoT protocols) but inherits, extends,
  or enhances them, providing core capabilities such as unified entity
  identity, physical/social/production relations graphs, and physical
  constraint integration. A2A is the most suitable starting-point when
  designing RVP.

  A2A Protocol Enhancement: Introduce RVP's core concepts into the A2A
  protocol to support physical-digital unified coordination.

    Extend Agent Card: Add composite identity description, dynamic
  capability discovery, and physical/social/production relations graph
  information to A2A's Agent Card to support intelligent routing.

    Unified Identity Mapping: Extend agent identity in A2A to unified
  entity identity (including Real-entity agent and  Virtual-entity
  agent), achieving cross-domain identity consistency.

    Real-time Coordination Semantics: Define new message types and
  interaction patterns for time-sensitive operations, supporting
  deadlines and real-time guarantees.

    Physical Constraint Representation: Add physical constraint
  fields to A2A messages, allowing agents to consider physical
  limitations (such as location, capabilities, time, etc.) during
  negotiation.



    zhang                  Expires 20,April,2026                [Page 15]

    Internet-Draft          RVP Problem Statement             October 2025


    Safety-aware Communication: Integrate safety context into A2A
  communication, including permission verification based on production
  relationships and physical safety constraints.

  Besides A2P, MCP and IoT protocols can also integrate with RVP, but
  this is not mandatory in the 1st phase:

  MCP Integration Enhancement: Enable MCP to support stateful, real-
  time coordination of physical agents.

  Hardware Tool Extension: Extend MCP tools from predominantly
  software digital domain to hardware physical domain, supporting
  embodied intelligence agent communication for hardware device
  queries, instructions, and interactions.

    Stateful Tool Representation: Encapsulate physical agents as
  stateful MCP tools, allowing tools to maintain state (such as device
  status, task progress).

    Timing Constraint Specification: Add timing constraints (such as
  start time, deadline, execution duration, etc.) in MCP tool calls.

    Multi-AI Coordination Support: Through RVP's
  physical/social/production relations graph, coordinate multiple AI
  model operations on the same physical system.

  IoT Protocol Bridge: Seamlessly integrate IoT devices into the RVP
  ecosystem, achieving both semantics of device data and intelligent
  coordination of device capabilities.

    Semantic Mapping: Map device data (such as sensor readings) to
  coordination context (such as events, states) and attach physical
  constraints.

    Unified Entity Identity: Assign unified entity identity to each
  IoT device and associate it with the corresponding Real-entity
  agents.

    Device Capability Representation: Describe device capabilities as
  standardized services, including functions, constraints, and states.

    Message Delivery Guarantees: Provide reliability guarantees (such
  as at-least-once delivery, acknowledgment mechanisms) for
  coordination messages among IoT devices.





    zhang                  Expires 20,April,2026                [Page 16]

    Internet-Draft          RVP Problem Statement             October 2025


5. RVP Use Cases

  This section is not a to-do list for the protocol. To make RVP well
  understood, it provides several scenarios on how RVP could be used
  in practice.

5.1. Flexible Production Line Coordination in Manufacturing

  A flexible manufacturing line produces customized products with
  different kinds of agents among robots, CNC machines, AGVs
  (Automated Guided Vehicles), and quality inspection stations, along
  with their digital twins [DIGITAL-TWINS] agents.

  A robot arm Composite Agent is composed of the controller, Real-
  entity robot arm Agent and its digital twins (Virtual-entity Agent).
  When the composite robot arm agent is registered via RVP, the
  associated real-entity robot and virtual-entity agents are
  automatically registered as well, reducing the complexity of
  registration several times.

  The Physical/Social/Production Relations Graph with hierarchical
  structure is created and registered which may look like: Production
  Line Manager Work Cell Controller Individual Robots. Permission
  constraints ensure only authorized controllers can issue commands to
  specific robots.

  Physical Constraints are transferred via RVP with spatial
  constraints such as robot workspace boundaries, collision avoidance
  zones; temporal constraints such as task completion deadlines and
  synchronization requirements; and dynamic constraints such as
  Maximum acceleration, payload capacity, energy consumption, etc....

  When the order is received, the production line manager agent
  decomposes tasks based on Physical/Social/Production Relations Graph
  and message routing is based on the relations graph, supporting
  explicit routing paths and automatic path calculation. The tasks in
  associated agents are linked based on the relations graph
  constraints supporting synchronous and asynchronous communication.
  The real-entity agents execute the appointed tasks with real-time
  monitoring and the virtual-entity agents simulate execution
  feasibility with physical constraints. The physical verification
  ensures digital instructions comply with physical system limitations
  after the tasks are completed. Constraint violation triggers
  automatic redo. The feedback as well as the process data in
  different agents can be integrated to the owner for decision-making
  and model training. Problematic nodes request assistance via
  relations graph queries using RVP.


    zhang                  Expires 20,April,2026                [Page 17]

    Internet-Draft          RVP Problem Statement             October 2025


5.2. Disaster Rescue Human-Robot Collaborative Environments

  Consider a search and rescue operation with human rescuers, rescue
  robots, drones, and command center.

  Robots and humans form ad-hoc coordination network in
  infrastructure-degraded environment. Flexible social relations graph
  adapting to situation is created via RVP to establish command
  structure including incident commander, who provides overall
  coordination; team leaders who coordinate local resources and
  Individual robots/humans, who have autonomy for immediate safety
  decisions. The social relations graph will dynamically change to add
  emergency relations (e.g., fire truck and ambulance).

  The robot Composite Agents register their capabilities via RVP and
  the search area is decomposed based on the robot capabilities, human
  responsibility and environmental constraints to create several
  search teams(including humans and robots). The physical constraint
  is monitored for structural integrity or hazardous conditions. The
  virtual-entity agents simulate evacuation routes with physical
  constraint verification. Once the constraint violation is detected,
  the digital twin agent changes its simulation and automatic
  evacuation action is executed via the social relations graph.
  Dynamic re-planning and re-verification is executed as new
  constraints are discovered (e.g., aftershock occurs or blocked route)
  for the real and virtual agents. Real-time data from operation feeds
  embodied disaster response model.

5.3. Autonomous Driving Systems in Smart City

5.3.1. Multi-Vehicles(Agents) Collaborative Delivery

  The fleet of autonomous delivery robots and drones across urban
  environment with dynamic obstacle avoidance and resource sharing is
  a multi-agents coordination. Each delivery vehicle is a Composite
  Agent including: high-level mission planner, the physical vehicle
  with sensors/actuators and virtual-entity agent for route
  optimization and simulation. The production Relations graph is like:
  Fleet management center coordinating individual vehicles peer
  relations between vehicles for resource negotiation hierarchical
  override for emergency situations.

  The virtual-entity agent simulates route with physical feasibility
  check. The production relations graph queries nearby vehicles for
  resource sharing opportunities and coordination sessions are
  established. The real-entity agent executes with continuous
  constraint monitoring and trigger dynamic re-planning when when the


    zhang                  Expires 20,April,2026                [Page 18]

    Internet-Draft          RVP Problem Statement             October 2025


  constraint violates (e.g., unexpected obstacle, battery drain,
  etc...). The execution data (actual vs. predicted energy consumption,
  obstacle patterns) feeds the autonomous driving model via RVP.

 5.3.2. Vehicle-to-Everything (V2X) Intersection Management

  Consider an intelligent intersection with autonomous V2X vehicles,
  human-driven V2X vehicles, cyclists, and pedestrians without
  traditional traffic signals.

  Once the vehicle approaches intersection, it registers it vehicle
  capabilities (e.g., size, braking distance, speed range and V2X
  information) and discovers intersection controller via RVP. The
  intersection controller establishes coordination sessions based on
  the physical relations graph among the vehicles. The virtual-entity
  agents simulate safe crossing sequences with physical constraint
  verification where physical safety constraints are the primary
  protocol feature to be checked, not afterthought. The Safety
  constraints like minimum time gaps, line-of-sight requirements,
  pedestrian priority are checked. The V2X vehicle, cyclist,
  pedestrian with smartphone receive crossing permissions with timing
  constraints via RVP. Continuous safety monitoring is executed and
  constraint violations trigger RVP protocols usage.

6. Security Considerations

6.1. Identity Verification and Authorization

  Composite Identity Verification Not only verify digital identity but
  also verify physical identity, ensuring authenticity of unified
  identity.

    Digital Identity: OAuth 2.0 tokens, API keys

    Physical Identity: Device fingerprinting, TPM attestation, secure
  elements

    Composite Binding: Cryptographic binding between physical and
  digital credentials

  Multi-factor Authentication for critical operations, require multi-
  factor authentication, including physical tokens or biometrics
  tokens:

    Level 1 (routine operations): Single factor (token/certificate)

    Level 2 (sensitive operations): Two factors (token + OTP)


    zhang                  Expires 20,April,2026                [Page 19]

    Internet-Draft          RVP Problem Statement             October 2025


    Level 3 (critical operations): Three factors (certificate +
  biometric + physical token)

  Authorization Based on Relations Graph

   Utilize physical/social/production relations graph for dynamic
   permission calculation, considering relations type, path length,
   time validity, etc.

6.2. Communication Security

  End-to-End Encryption

  Use TLS 1.3 or higher for transport encryption, and apply
  application-layer end-to-end encryption for sensitive data.

  Message Integrity

  Use digital signatures to ensure message integrity and prevent
  tampering.

6.3. Safety System Protection

  Independent Safety Channels

  Physical safety systems (such as emergency stop) use communication
  channels independent of coordination protocols.

  Physical-Digital Boundary Protection

  Deploy safety gateways between physical and digital domains for
  protocol filtering and deep inspection.

  Gateway functions include:

    Message content inspection and validation

    Rate limiting and anomaly detection

    Whitelist-based operation filtering

    Constraint verification before physical execution

6.4. Audit and Monitoring

  Comprehensive Logging Record: Detailed logs of all coordination
  sessions, including message content, participants, timestamps, etc.


    zhang                  Expires 20,April,2026                [Page 20]

    Internet-Draft          RVP Problem Statement             October 2025


6.5. Isolation Capabilities

  Dynamic Isolation: Based on security events, dynamically isolate
  affected entities or coordination domains.

7. IANA Considerations

  This document requests IANA to establish new registries for RVP
  protocol parameters. More details are in further versions.

8. References

8.1. Normative References

  [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
            Requirement Levels", BCP 14, RFC 2119, March 1997.

8.2. Informative References

  [A2A-SPEC] https://a2a-protocol.org/dev/specification/

  [A2A-FIPA] Foundation for Intelligent Physical Agents, "FIPA Agent
              Communication Language Specifications", 2002.

  [MCP-SPEC] Anthropic, "Model Context Protocol Specification", 2024.

  [ANP-SPEC] "Agent Network Protocol Technical White Paper",
              https://arxiv.org/pdf/2508.00007, 2025.

  [RFC7252] Shelby, Z., Hartke, K., and C. Bormann, "The Constrained
            Application Protocol (CoAP)", RFC 7252, June 2014.

  [SMART-FACTORY] https://www.globaltrademag.com/smart-factories-how-
                   technology-is-revolutionizing-manufacturing/.

  [AI-CITY] https://www.deloitte.com/content/dam/assets-
            shared/docs/industries/government-public-services/2025/ai-
            powered-cities-of-the-future.pdf.

  [APOLLO]  https://www.apollo.auto/en/apollo-self-driving.

  [Human-Robot] Dong, W., "Toward Embodied Intelligence-Enabled Human-
            Robot Symbiotic Manufacturing: A Large Language Model-
            Based Perspective",Journal of Computing and Information
            Science in Engineering, May 2025.




    zhang                  Expires 20,April,2026                [Page 21]

    Internet-Draft          RVP Problem Statement             October 2025


  [DIGITAL-TWINS] Grieves, M., "Digital Twin: Manufacturing Excellence
            through Virtual Factory Replication", Digital
            Manufacturing, 2014.

  [CPS-DESIGN] Lee, E., "Cyber Physical Systems: Design Challenges",
            University of California, Berkeley Technical Report, 2008.

  [STI-BENCH] Li, Y., "STI-Bench:Are MLLMs Ready for Precise Spatial-
            Temporal World Understanding?", ICCV 2025.

  [DATA-RUNOUT] Villalobos,P.,Will we run out of data? Limits of LLM
            scaling based on human-generated data,
            https://arxiv.org/pdf/2211.04325.



9. Acknowledgments

  The authors thank Dirk Dressler who provided valuable feedback
  during the development of this specification.




























    zhang                  Expires 20,April,2026                [Page 22]

    Internet-Draft          RVP Problem Statement             October 2025


    Authors' Addresses


  Xiaoxun Zhang
  Hicagent
  No.57 Boxia Road, Pudong Software Park,Shanghai, P.R.China
  zhangxiaoxun@hicagent.com

  Yunfei Zhang(Editor)
  China Telecom
  No.31 Jinrong Street, Xicheng District,Beijing, P.R.China
  Zhangyf80@chinatelecom.cn

  Chi Zhang
  China Telecom
  No.31 Jinrong Street, Xicheng District,Beijing, P.R.China
  Zhangc120@chinatelecom.cn

  Min Zhang
  China Telecom
  No.31 Jinrong Street, Xicheng District,Beijing, P.R.China
  Zhangmin3@chinatelecom.cn


  Weijun Lu
  SUPCON  TECHNOLOGY  CO., LTD.
  No.309 Liuhe Road, BinJiang District, Hangzhou,Zhejiang, P.R.China
  luwj@supcon.com




















    zhang                  Expires 20,April,2026                [Page 23]

