Digital Twin-Enabled Online Battlefield Learning with Random Finite Sets

Area of Science:

• Robotics and Autonomous Systems
• Artificial Intelligence
• Defense Technology

Background:

• Digital twin technology is a key emerging innovation for unmanned combat systems.
• Online battlefield learning is crucial for effective digital twin implementation in real-time combat scenarios.
• Existing methods struggle with real-time battlefield state learning due to detection uncertainties.

Purpose of the Study:

• To propose a novel random finite set- (RFS-) based algorithm for online battlefield learning.
• To enable real-time battlefield state estimation for digital twin applications in unmanned combat.
• To address challenges posed by detection uncertainties, including clutter, missed detections, and noise.

Main Methods:

• Development of an RFS-based digital twin framework including battlefield state, UGV motion, and sensor models.
• Modification of the probability hypothesis density (PHD) filter using Bayesian inference for online learning.
• System architecture and operational modes for digital twin-enabled online learning are detailed.

Main Results:

• The proposed algorithm effectively performs online learning of battlefield states in the presence of uncertainties.
• Experimental validation using an unmanned ground vehicle (UGV) demonstrates the algorithm's performance.
• The system architecture and operational modes are systematically described and validated.

Conclusions:

• The developed RFS-based algorithm provides an effective solution for real-time battlefield learning in digital twin-enabled unmanned combat.
• This research demonstrates the practical application and effectiveness of digital twin technology in enhancing unmanned combat capabilities.
• The findings offer a valuable framework for future advancements in autonomous combat systems.