Extending the Framework for Developing Intelligent Virtual Environments (FIVE) with Artifacts for Modeling Internet...

Area of Science:

• Distributed Systems
• Artificial Intelligence
• Robotics

Background:

• Federated Learning (FL) is a key area in distributed machine learning.
• Wireless Ad-hoc Networks (WANETs) present unique challenges for agent communication and model sharing.
• Ensuring agent network connectivity is vital before deploying autonomous systems.

Purpose of the Study:

• To apply a decentralized Federated Learning algorithm based on consensus (CoL) to WANETs.
• To evaluate agent reachability in WANETs through realistic simulations.
• To introduce the concept of 'artifacts' within a simulation framework to model critical devices.

Main Methods:

• Implementation of a consensus-based Federated Learning algorithm (CoL) in a WANET environment.
• Development of a realistic simulation framework (FIVE) using Unity and SPADE agents.
• Case study involving autonomous tractors in a fruit orchard, incorporating 'artifacts' for connectivity.

Main Results:

• Demonstration of CoL algorithm's applicability in WANETs.
• Validation of the FIVE framework for simulating agent interactions and network connectivity.
• Successful integration of 'artifacts' to represent essential, location-specific devices.

Conclusions:

• The CoL algorithm is suitable for decentralized learning in WANETs.
• The FIVE framework provides a robust platform for simulating complex agent-based systems.
• The 'artifact' concept is a foundational step towards modeling Digital Twins and Digital Shadows.