Self-configurable Manufacturing Industrial Agents: SMIA

The Self-configurable Manufacturing Industrial Agents (SMIA) is a proposal for the implementation of the concept of the I4.0 Component from the Reference Architectural Model Industrie 4.0 (RAMI 4.0) as an AAS-compliant agent-based Digital Twin (DT). The features of the SMIA approach include:

• free & open-source
• AAS-compliant: standardized approach
• Ontology-based
• easily customizable and configurable
• self-configuration at software startup
• easy to usage
• containerized solution

💡 TIP: For more details on Self-configurable Manufacturing Industrial Agents (SMIA) see the 📄 full documentation.

The SMIA approach originated as part of a PhD thesis addressed by Ekaitz Hurtado in the Systems Control and Integration (GCIS) research group at the University of the Basque Country (EHU). Following the completion of the thesis, the framework remains under active and continuous development to expand its ecosystem and enhance its industrial maturity.

ℹ️ IMPORTANT: The SMIA software is developed as part of scientific research. If you use SMIA in scientific works, please cite the following articles.

Scientific contribution

E. Hurtado, A. Burgos, A. Armentia, and O. Casquero, Self-configurable Manufacturing Industrial Agents (SMIA): a standardized approach for digitizing manufacturing assets, Journal of Industrial Information Integration, vol. 47, p. 100915, Sept. 2025, doi: 10.1016/j.jii.2025.100915

Software contribution

E. Hurtado, I. Sarachaga, A. Armentia, and O. Casquero, An open-source reference framework for the implementation of type 3 Asset Administration Shells, Software Impacts, vol. 27, p. 100807, Apr. 2026, doi: 10.1016/j.simpa.2025.100807

Usage

❗ IMPORTANT: The SMIA reference framework is under active and continuous development, regularly introducing new features, architectural improvements, and bug fixes. Ongoing experimental validations and new industrial use cases continue to be integrated as the ecosystem evolves.

Multiple ways of running SMIA software are available. The associated GitHub repository shows how to run the base SMIA software.

In this case, how to use the SMIA Python package is shown.

Installation

The SMIA approach can be easily installed using pip:

pip install smia

Available facilities

The SMIA approach offers different facilities for its use:

• smia.launchers: some launchers to run the base SMIA software.
  • smia.launchers.smia_cli_starter.py: launcher to run the SMIA software by passing the AAS model as CLI argument.
  • smia.launchers.smia_starter.py: launcher to run the SMIA software by passing the AAS model through Python code.
  • smia.launchers.smia_docker_starter.py: launcher to run the SMIA software as Docker container by passing the AAS model as environmental variable.
• smia.agents: the Agent classes available to be instantiated and used.
  • smia.agents.SMIAAgent: the generic SMIA agent.
  • smia.agents.ExtensibleSMIAAgent: the extensible SMIA agent offering all extension methods.
• smia.assetconnection: tools related to connection with assets.
  • smia.assetconnection.HTTPAssetConnection: tools related to connection with assets through HTTP communication protocol.

The other modules are available for import when, for example, developing an extension to SMIA.

Extensibility

SMIA extensibility is provided through the special agent ExtensibleSMIAAgent. It provides some methods to extend the base SMIA and add own code:

• add_new_agent_capability(behaviour_class): this method adds a new agent capability to SMIA to increase its intelligence and autonomy. The new capability is added as a SPADE behavior instance.
• add_new_agent_service(service_id, service_method): this method adds a new agent service to SMIA to increase its intelligence and autonomy. The new service is added as a Python method that will be called when the service is requested.
• add_new_asset_connection(aas_interface_id_short, asset_connection_class): this method adds a new asset connection to SMIA. The new connection is added by the instance class inherited from the official SMIA generic class AssetConnection and the associated AAS interface element.

Examples

Create and run an SMIAAgent:

import smia
from smia.agents.smia_agent import SMIAAgent

smia.load_aas_model('<path to the AASX package containing the AAS model>')
my_agent = SMIAAgent()

smia.run(my_agent)

Extensibility examples

Create an ExtensibleSMIAAgent:

import smia
from smia.agents.extensible_smia_agent import ExtensibleSMIAAgent

smia.load_aas_model('<path to the AASX package containing the AAS model>')
my_agent = ExtensibleSMIAAgent()

Add new Agent Capability to ExtensibleSMIAAgent:

import spade
from smia.agents.extensible_smia_agent import ExtensibleSMIAAgent  # Import from the SMIA package

new_capability = spade.behaviour
my_extensible_agent.add_new_agent_capability(new_capability)

Add new Agent Service to ExtensibleSMIAAgent:

from smia.agents.extensible_smia_agent import ExtensibleSMIAAgent  # Import from the SMIA package

my_extensible_agent.add_new_agent_service(new_service_id, new_service_method)

Add new Asset Connection to ExtensibleSMIAAgent:

from smia.agents.extensible_smia_agent import ExtensibleSMIAAgent  # Import from the SMIA package

my_extensible_agent.add_new_asset_connection(aas_interface_id_short, asset_connection_class)

Complete example of an Extensible SMIA agent:

import smia
import asyncio
from spade.behaviour import CyclicBehaviour
from smia.agents.extensible_smia_agent import ExtensibleSMIAAgent  # Import from the SMIA package
from my_asset_connections import MyAssetConnection

class MyBehaviour(CyclicBehaviour):
    async def on_start(self):
        print("MyBehaviour started")
        self.iteration = 0

    async def run(self):
        print("MyBehaviour is in iteration {}.".format(self.iteration))
        self.iteration += 1
        await asyncio.sleep(2)

def new_service_method():
    print("This is a new service to be added to SMIA.")
    
def main():
    my_extensible_agent = ExtensibleSMIAAgent()
    smia.load_aas_model('<path to the AAS model>')
    
    my_behav = MyBehaviour()
    my_extensible_agent.add_new_agent_capability(my_behav)
    
    my_extensible_agent.add_new_agent_service('new_service_id', new_service_method)
    
    new_asset_connection = MyAssetConnection()
    my_extensible_agent.add_new_asset_connection('new_connection_idshort', new_asset_connection)
    
    smia.run(my_extensible_agent)

if __name__ == '__main__':
    main()

Dependencies

The SMIA software requires the following Python packages to be installed to run correctly. These dependencies are listed in pyproject.toml so that they are automatically obtained when installing with pip:

• spade (MIT license)
• basyx-python-sdk (MIT license)
• owlready2 (GNU LGPL licence v3)

License

GNU General Public License v3.0. See LICENSE for more information.